Category: PowerScale & Isilon

OneFS SmartQuotas help measure, predict, control, and limit the rate of storage capacity consumption, allowing precise cluster provisioning to best meet an organization’s storage needs. SmartQuotas also enables ‘thin provisioning’, or the ability to present more storage capacity to applications and users than is physically present (over-provisioning). This allows storage capacity to be purchased and provisioned organically and in real time, rather than making large, speculative buying decisions ahead of time. As we will see, OneFS also leverages quotas for calculating and reporting on data reduction and storage efficiency across user-defined subsets of the /ifs file system.

SmartQuotas provides two fundamental types of capacity quota:

• Accounting Quotas
• Enforcement Quotas

Accounting Quotas simply monitor and report on the amount of storage consumed, but do not take any limiting action or intervention. Instead, they are primarily used for auditing, planning, or billing purposes. For example, SmartQuotas accounting quotas can be used to:

• Generate reports to analyze and identify storage usage patterns and trends. These can then be used to define storage policies for the business.
• Track the amount of disk space used by various users, groups, or departments to bill each entity for only the storage capacity they actually consume (charge-back).
• Intelligently plan for capacity expansions and future storage need.

The ‘isi quota quotas create –enforced=false’ CLI command can be used to create an accounting quota. Alternatively, this can be done from the WebUI by navigating to File System > SmartQuotas > Quotas and usage > Create quota.

The following CLI command creates an accounting quota for the /ifs/data/acct_quota_1 directory, setting an advisory threshold that is informative rather than enforced.

# isi quota quotas create /ifs/data/acct_quota_1 directory \ --advisory-threshold=10M --enforced=false

Before using quota data for analysis or other purposes, verify that no QuotaScan jobs are in progress by running the following CLI command:

# isi job events list --job-type quotascan

In contrast, enforcement quotas include all of the functionality of the accounting option plus the ability to limit disk storage and send notifications. The ‘isi quota quotas create –enforced=true’ CLI syntax can be used to create an enforcement quota. Alternatively, this can be done from the WebUI by navigating to File System > SmartQuotas > Quotas and usage > Create quota.

The following CLI command creates an enforcement quota for the /ifs/data/enforce_quota_1 directory, setting an advisory threshold that is informative rather than enforced.

# isi quota quotas create /ifs/data/enforce_quota_1 directory \ --advisory-threshold=10M --enforced=true

Using enforcement limits, a cluster can be logically partitioned in order to control or restrict how much storage that a user, group, or directory can use. For example, capacity limits can be configured to ensure that adequate space is always available for key projects and critical applications – and to ensure that users of the cluster do not exceed their allotted storage capacity.

Optionally, real-time email quota notifications can be sent to users, group managers, or administrators when they are approaching or have exceeded a quota limit.

A OneFS quota can have one of four enforcement types:

Enforcement	Description
Hard	A limit that cannot be exceeded. If an operation such as a file write causes a quota target to exceed a hard quota, the operation fails, an alert is logged to the cluster and a notification is sent to any specified recipients. Writes resume when the usage falls below the threshold.
Soft	A limit that can be exceeded until a grace period has expired. When a soft quota is exceeded, an alert is logged to the cluster and a notification is issued to any specified recipients. However, data writes are permitted during the grace period. If the soft threshold is still exceeded when the period expires, writes will be blocked, and a hard-limit notification issued to any specified recipients.
Advisory	An informal limit that can be exceeded. When an advisory quota threshold is exceeded, an alert is logged to the cluster and a notification is issued to any specified recipients. Reaching an advisory quota threshold does not prevent data writes.
None	No enforcement. Quota is accounting only.

All three quota types have both a limit, or threshold, and a grace period. In OneFS 8.2 and later, a soft quota and advisory quota threshold can be specified as a percentage, as well as a specific capacity. For example:

# isi quota quotas create /ifs/quota directory --percent-advisory-threshold=80 --percent-soft-threshold=90 --soft-grace=1d --hard-threshold=100G

A hard quota has a zero-time grace period, an advisory quota has an infinite grace period and a soft quota has a configurable grace period. When a quota limit and grace period have been exceeded, a client write operations to anywhere within that quota domain will fail with EDQUOT. Although enforcements are implemented generically in the quota data bases, only one resource may be limited per domain, either logical or physical space.

Even when a hard quota limit is reached, there are certain instances where operations are not blocked. These include administrative control through root (UID 0), system maintenance activities, and the ability of a blocked user to free up space.

The table below describes the three SmartQuotas enforcement states:

Enforcement State	Description
Under (U)	If the usage is less than the enforcement threshold, the enforcement is in state U.
Over (O)	If the usage is greater than the enforcement threshold, the enforcement is in state O.
Expired (E)	If the usage is greater than the soft threshold, and the usage has remained over the enforcement threshold past the grace period expiration, the soft threshold is in state E. If an administrator modifies the soft threshold but not the grace period, and the usage still exceeds the threshold, the enforcement is in state E.

There are a few exceptions to enforcement of Quotas including the following scenarios:

• If a domain has an accounting only quota, enforcements for the domain are not applied.
• Any administrator action may push a domain over quota. Examples include changing protection, taking a snapshot, or removing a snapshot. The administrator may write into any domain without obeying enforcements.
• Any system action may push a domain over quota, including repair. OneFS maintenance processes are as powerful as the administrator.

Governance is the mechanism by which SmartQuotas determines which domains apply to a given file or directory. After a sequence of domain configuration changes, a persistent record is needed in order to know where a file had been accounted. As such, quotas utilize ‘tagging’, and the governing domains are recorded in a dynamic attribute of the inode.

A Quota Domain Account tracks usages and limits of a particular domain. For scalability reasons, the QDA system dynamically breaks up the Quota Domain’s account of the quota into some number of Quota Domain Account Constituents (QAC), each of which tracks a part of the account. Modifications to the account are distributed at random among these constituents. Each Quota Domain Account Constituent is stored in a set of mirrored Quota Accounting Blocks (QABs). QABs track usage of a quota and consist of several level counters for different tracked resource types and level limits for advisory, soft, and hard quotas.

The Quota Domain Record stores all configuration and state associated with a domain. The record can be subdivided into three components:

Component	Description
Configuration	Quota configuration.
Enforcement	This includes the grace period, limit, and notification state.
Account	The mechanism for space utilization accounting.

With SmartQuotas, there are three main ways of tracking, enforcing, and reporting resource usage:

Tracking Method	Description
Physical size	This is simple to track, since it includes all the data and metadata resources used, including the data-protection overhead. The quota system is also able to track the difference before and after the operation.
File system logical size	This is slightly more complex to calculate and track but provides the user with a more comprehensible means of understanding their usage.
File accounting	This is the most straightforward, since whenever a file is added to a domain, the file count is incremented.
Application logical size	Reports total logical data store across different tiers, including CloudPools, to account for the exact file sizes. Allows users to view quotas and free space as an application would view it, in terms of how much capacity is available to store logical data, regardless of data reduction or tiering technology.

Prior to OneFS 8.2, SmartQuota size accounting metrics typically used a count of the number of 8 KB blocks required to store file data on cluster. Accounting based on block count can result in challenges, such as small file over-reporting. For example, a 4KB file would be logically accounted for as 8KB. Similarly, block-based quota accounting only extends to on-premises capacity consumption. This means that a 100MB file stored within a CloudPools tier would only be account for as an 8KB SmartLink stub file, rather than its actual size.

To directly address this issue in OneFS 8.2 and later, application logical quotas provide an additional quota accounting metric. Application logical size accounts for, reports, and enforces on the actual space consumed and available for storage, independent of whether files are cloud-tiered, sparse, deduplicated, or compressed. Application logical quotas can be easily configured from the CLI with the following syntax:

# isi quota quotas create <dir> directory –-thresholds-on=applogicalsize

Any legacy quotas created on OneFS versions prior to 8.2 can easily be converted to use application logical size upon upgrade.

For logical space accounting, some inode attributes such as ACLs and symbolic links are included in the resource count. This uses the same data that is displayed in the ‘logical size’ field by the ‘isi get –DD <file>’ CLI command.

As we saw in the previous article in this series, SmartPools operation is quarterbacked and executed by the OneFS job engine.

When a one of the SmartPools job runs, all the files’ attributes are examined and checked against the list of file pool policies. As such, file pool policies are built on file attribute(s) the policy can match on, and these include file name, path, file type, size, timestamps, etc.

Once the file attribute is set to select the appropriate files, the action to be taken on those files can be added. For example, if the selected attribute is File Size, additional settings are available to dictate thresholds – for instance, all files bigger than 500MB, but smaller than 2GB. Next, actions are applied, such as move to node pool ‘x’, protect at level ‘y’, and lay out for access setting ‘z’.

File Attribute	Description
File Name	Specifies file criteria based on the file name
Path	Specifies file criteria based on where the file is stored
File Type	Specifies file criteria based on the file-system object type
File Size	Specifies file criteria based on the file size
Modified Time	Specifies file criteria based on when the file was last modified
Create Time	Specifies file criteria based on when the file was created
Metadata Change Time	Specifies file criteria based on when the file metadata was last modified
Access Time	Specifies file criteria based on when the file was last accessed
User Attributes	Specifies file criteria based on custom  attributes – see below

Path-based file pool policies can direct data to the correct node pool on write, without a SmartPools job running.  However, policies that use other attributes beside path to dictate placement, move their matching data when the next SmartPools job runs. This ensures that write performance is not sacrificed for initial data placement. Data not covered by a file pool policy is targeted to the default tier, which can be configured as desired. Note that CloudPools, the OneFS off-cluster cloud tiering service, also uses the file pool policy engine.

File pool policies can be configured from the CLI using the ‘isi filepools create’ command, or via the WebUI by navigating to File System > Storage Pools > File Pool Policies > Create a file pool policy:

When a file pool policy is created, SmartPools stores it in a configuration database with any other file policies.  When a SmartPools job runs, it applies all the policies in order.  If a file matches multiple policies, SmartPools will only apply the first rule it matches.  So, for example if there is a rule that moves all small jpeg files to an A-series archive pool, and another that moves all files under 1 MB to an F-series performance tier, if the jpeg rule appears first in the list, then jpg files under 2 MB will go to archive, NOT the performance tier.

Criteria can be combined within a single policy using ‘And’ or ‘Or’ operators, so that data can be classified very granularly.  Continuing with our example, if the desired behavior is to have all jpg files over 2 MB to be moved to the Archive node pool, the file pool policy can be simply constructed with an ‘And’ operator to explicitly cover that condition.

While the example above is a simple one, if needed, SmartPools can currently support up to 128 file pool policies, each of which can contain up to 3 file matching criteria or rules. However, as the list of file pool policies grows large, it becomes less practical to manually traverse them to see how a file will behave when policies are applied.

File pool policy order, and policies themselves, can be easily changed at any time. Specifically, policies can be added deleted, edited, copied and re-ordered. For example:

# isi filepool policies modify Archive_1 --description "Move older files to archive storage" --data-storage-target Archive_1 --data-ssd-strategy metadata --begin-filter --file-type=file --and --birth-time=2022-10-01 --operator=lt --and --accessed-time=2022-11-01 --operator=lt --end-filter

The file pool policy is applied when the next scheduled SmartPools job runs. By default, the SmartPools job runs once a day, but can also started manually:

# isi job jobs start SmartPools

File pool policies are evaluated in descending order, according to their position in the file pool policies list. By default, when a new policy is created, it is inserted immediately above the default file pool policy. The default policy is always the last in priority, and applies to all files that are not matched by any other file pool policy. The priority order of a file pool policy can be altered by moving it up or down in the list. For example:

# isi filepool policies list

Name        Description                               CloudPools State

----------------------------------------------------------------

Archive_1   Move older files to archive storage       No access

Perf_1      Move recent files to perf tier            No access

----------------------------------------------------------------

Total: 2

# isi filepool policies modify Perf_1 --apply-order 1

# isi filepool policies list

Name        Description                               CloudPools State

----------------------------------------------------------------

Perf_1      Move recent files to perf tier            No access

Archive_1   Move older files to archive storage       No access

----------------------------------------------------------------

Total: 2

In this case, the ‘Perf_1’ policy has been promoted to the top of the list, above the ‘Archive_1’ policy.

If no File Pool policy matches a file, the default policy specifies all storage settings for the file. The default policy, in effect, matches all files not matched by any other SmartPools policy. For this reason, the default policy is the last in the file pool policy list, and, as such, always the last policy that SmartPools applies.

Additionally, a file pool policy can be configured to match a user-specified ‘custom attribute’ and/or value.

When data is written to the cluster, SmartPools writes it to a single Node Pool only.  This means that, in almost all cases, a file exists in its entirety within a Node Pool, and not across Node Pools.  SmartPools determines which pool to write to as follows:

• If a file matches a file pool policy based on directory path, that file will be written into the Node Pool dictated by the File Pool policy immediately.
• If a file matches a file pool policy which is based on any other criteria besides path name, SmartPools will write that file to the Node Pool with the most available capacity.

The OneFS ‘isi get –D’ CLI command, or WebUI File System Explorer, provides a detailed view of where SmartPools-managed data is at any time by both the actual Node Pool location and the File Pool policy-dictated location (i.e. where that file will move after the next successful completion of the SmartPools job). More specifically, the selection of a disk pool target from a file pool policy typically follows the following logic path:

1. If SmartPools is licensed and the policy’s pool ID is found that disk pool is targeted.
2. If SmartPools is unlicensed, the policy ID specified for a file is ignored and the ‘any disk pool’ group ID is used instead.
3. If the policy ID is not found and global spillover is enabled for the cluster, the spillover target is used as the policy. If global spillover is disabled, the ‘any disk pool’ group is used as the policy.
4. The pools in the policy which satisfy the SSD preference are presented in a weighted random order. This continues until a suitable pool is found or an error is returned.
5. If no suitable pool is found, the SSD preference is changed to ‘fallback’ and step #3 is repeated. The ‘fallback’ value allows the use of any pool if the reserved ‘system’ policy is used, or the cluster is all-SSD. Otherwise only all-HDD pools are used.
6. If no suitable pool is found and global spillover is enabled for the cluster, the spillover target is used as the policy and step #3 is repeated (the SSD preference remains as ‘fallback’).
7. If spillover is disabled and no suitable pool is found, processing stop and an error is returned.

After a file match with a File Pool policy occurs, the OneFS uses the settings in the matching policy to store and protect the file. However, a matching policy might not specify all settings for the match file. In this case, the default policy is used for those settings not specified in the custom policy. For each file stored on a cluster, the system needs to determine the following:

·         Requested protection level

·         Data storage target for local data cache

·         SSD strategy for metadata and data

·         Protection level for local data cache

·         Configuration for snapshots

·         SmartCache setting

·         L3 cache setting

·         Data access pattern

·         CloudPools actions (if any)

A question that’s frequently asked is what happens to any files that are due to be tiered but are being actively used? SmartPools can move the files transparently, even if they’re open and being modified.

Under the hood, the locks OneFS uses to provide consistency inside the filesystem, are separate from the external file locks for consistency between applications. This allows OneFS to discreetly move metadata and data blocks around, while the file is locked by an application. The restriper also performs its work in small chunks to further minimize disruption.

In addition to actual file placement, SmartPools data access (DAC) settings can be configured at the file pool, or even the single file, level for the type of application or workflow. DAC allows data to be optimized for concurrent, streaming or random access, with each of these three options influencing how files are laid out on disk and cached. Specifically, the ‘random’ data access setting performs little to no read-cache prefetching, to avoid wasted disk seeks. This works best for small files under 128KB, and large files with random, small block accesses. Data is striped across the minimum number of drives needed to achieve the data protection settings.

Streaming access works well for sequentially-read, medium to large files. This access pattern uses aggressive prefetching to improve overall read throughput, and on disk layout spreads the file across a large number of drives to optimize access.

Concurrency, the default, is the middle ground option with moderate prefetching, and data striped across the minimum number of drives required to achieve the configured protection setting. Concurrency is useful for general workloads like file shares and home directories, and file sets with a mix of both random and sequential access.

All the current generation of PowerScale nodes contain some percentage of flash media, and these SSDs can be used to accelerate performance across the entire cluster, by using them for caching or storage. As such, OneFS offers several SSD Strategies, including:

SSD Strategy	Description
Metadata read acceleration	Creates a preferred mirror of file metadata on SSD, and writes the rest of the metadata, plus all the actual file data, to HDD.
Metadata read & write acceleration	All the metadata mirrors are stored on SSD.
Avoid SSDs	Writes all associated file data and metadata to HDDs. Only really used when there is insufficient SSD storage capacity, to prioritize its utilization.
Data on SSDs	All of a node pool’s data and metadata resides on flash.
L3 cache	All of a node pool’s SSDs are used for SmartFlash read caching.

When L3 caching is enabled, it consumes all the SSD capacity in a node pool and therefore cannot coexist with other SSD strategies.

In contrast to L3 cache, with the data on SSD strategy, only the files specifically targeted to SSD benefit from the increased read and write performance. The remainder of the data on the node pool lives exclusively on hard disk and will not benefit from SSD.

The ‘isi_cache_stats -v’ CLI command will return the ratio of L3 cache hits to cache misses. A value of  70% or more cache hits indicates that L3 is working pretty well. Whereas less than 70% suggests that the SSDs may be better used for a metadata strategy.

However, be aware that SmartPools SSD strategies in general typically require more complex configuration than L3 and must be monitored so as not to exceed the available SSD capacity.

In summary, as far as good practices for optimal cluster performance, consider the following when deploying and configuring SmartPools:

• Define a performance and protection profile, or SLA, for each tier, and configure it accordingly.
• Avoid creating tiers that combine node pools with differing performance profiles (ie. with and without SSDs).
• Ensure that cluster capacity utilization, for both hard drives and SSDs, remains below 90%.
• Keep Virtual Hot Spares enabled, with a minimum of 10% space allocation.
• Avoid creating hardlinks to files which will cause the file to match different file pool policies
• If node pools are combined into tiers, craft file pool rules to target the tiers rather than individual node pools within the tiers.
• Determine if metadata operations for a particular workload are biased towards reads, writes, or an even mix, and select the optimal SmartPools metadata or L3 caching strategy.
• If attempting to configure ‘up-tiering’, ensure it does what you expect. SmartPools jobs are scheduled, so the promotion of a file from an archive to a performance tier will not be immediate upon its access or modification.
• When employing a deep archiving strategy, ensure that the performance pool is optimized for all directories and metadata and the archive tier just for cold files as they age. This can be configured by adding a ‘TYPE=FILE’ statement to the aging file pool policy rule(s) to only move files to the archive tier.
• If SmartPools takes more than a day to run, or the cluster is already running the FSAnalyze job, consider using the FilePolicy, and corresponding IndexUpdate job.
• When enabling and scheduling the FilePolicy job, continue running the SmartPools job at a reduced frequency. For example:
• IndexUpdate running every six hours (low impact and priority 5), FilePolicy running daily (low impact and priority 6), and the SmartPools job running on the first Sunday of each month ( low impact and priority 6).
• Use SmartPools for Painless Tech refresh with intra-cluster migrations of data to other node pools. Allowing data to drain from a node pool before decommissioning makes the SmartFail occur much faster.

And finally, a laudable mantra for SmartPools management could be “simplicity reigns”! Where possible, resist the temptation to create more tiers, policies, or rules (ie. complexity) than you actually need.

The previous article examined OneFS storage pools, the substrate upon which SmartPools data tiering is built.

Next up the stack are OneFS file pools – the SmartPools logic layer. User configurable file pool policies govern where data is placed, accessed, and protected, accessed, and how it moves among the node pools and tiers.

File pools allow data to be automatically moved from one type of storage to another within a single cluster, to meet performance, space, cost or other criteria – all while retaining its data protection settings, and without any stubs, indirection layers, or other file system modifications.

Under the hood, the OneFS job engine is responsible for enacting the file movement, as instructed by configured file pool policies.

In all, there are currently five job engine jobs associated with OneFS SmartPools:

Job	Description	Default Execution
SetProtectPlus	Applies the default file policy. This job is disabled if SmartPools is activated on the cluster	Daily @ 10pm if SP is unlicensed. Low impact, priority 6
SmartPools	Job that runs and moves data between the tiers of nodes within the same cluster. Also executes the CloudPools functionality if licensed and configured.	Daily @ 10pm Low impact, priority 6
SmartPoolsTree	Enforces SmartPools file policies on a subtree.	Manual Medium impact, priority 5
FilePolicy	Efficient changelist-based SmartPools file pool policy job.	Daily @ 10pm Low impact, priority 6
IndexUpdate	Creates and updates an efficient file system index for FilePolicy job.	Manual Low impact, priority 5

When SmartPools is unlicensed, any disk pool policies are ignored, and instead, the policy is considered to include all disk pools, and file data is directed to, and balanced across, all pools.

When a SmartPools job runs, it examines and compares file attributes against the list of file pool policy rules.  To minimize runtime, the initial scanning phase of the SmartPools’ job uses a LIN-based scan, rather than a more expensive tree-walk – and this is typically even more efficient when an SSD metadata acceleration strategy is used.

A SmartPools LIN tree scan breaks up the metadata into ranges for the cluster nodes to work on in parallel.  Each node can then dedicate multiple threads to execute the scan on their assigned range.  A LIN scan also ensures each file is opened only once, which is much more efficient when compared to a directory walk, where hard links and other constructs can result in single threading, multiple opens, etc.

When a file pool job thread finds a match between a file and a policy, it stops processing additional rules, since that match determines what will happen to the file. Next, SmartPools checks the file’s current settings against those the policy would assign, to identify those which do not match. Once SmartPools has the complete list of settings that it needs to apply to that file, it sets them all simultaneously, and moves to restripe that file to reflect any and all changes to node pool, protection, SmartCache use, layout, etc.

The file pool policy engine falls under the control and management of the SmartPools job. The default schedule for this process is every day at 10pm, and with a low impact policy. However, this schedule, priority and impact can be manually configured and tailored to a particular environment and workload.

SmartPools can also be run on-demand, to apply the appropriate file-pool membership settings to an individual file, or subdirectory, without having to wait for the background scan to do it.

For example, to test what affect a new policy will have, the ‘isi filepool apply’ command line utility can be run against a small subset of the data, which can be either a single file, or group of files or directories.  This CLI command can either be run live, to actually make the policy changes, or in a ‘dry-run’ assessment mode, using the ‘-nv’ flags, to estimate the scope and effect of a policy.

For a detailed view of where a SmartPools-managed file is at any time, the ‘isi get’ CLI command can provide both the actual node pool location, and the file pool policy-dictated location – or where that file will move to, after the next successful SmartPools job run.

When data is written to the cluster, SmartPools writes it to a single node pool only.  This means that, in almost all cases, a file exists in its entirety within a node pool, and not across pools

Unlike the SmartPools job, which scans the entire LIN tree, and the SmartPoolsTree job which visits a subtree of files, the FilePolicy job, introduced in OneFS 8.2, provides a faster, lower impact method for applying file pool policies. In conjunction with the IndexUpdate job, FilePolicy improves job scan performance, by using a snapshot delta based ‘file system index’, or changelist, to find files needing policy changes.

Avoiding a full treewalk dramatically decreases the amount of locking and metadata scanning work the job is required to perform, improving execution time, and reducing impact on CPU and disk – albeit at the expense of not quite doing everything that SmartPools does. However, most of the time SmartPools and FilePolicy perform the same work.  Disabled by default, FilePolicy supports a wide range of file policy features, reports the same information, and provides the same configuration options as the SmartPools job. Since FilePolicy is a changelist-based job, it performs best when run frequently – once or multiple times a day, depending on the configured file pool policies, data size and rate of change.

When enabling and using the FilePolicy and IndexUpdate jobs, the recommendation is to continue running the SmartPools job as well, but at a much-reduced frequency.

FilePolicy requires access to a current index. This means that if the IndexUpdate job has not yet been run, attempting to start the FilePolicy job will fail with an error message, prompting to run the IndexUpdate job first. And once the index has been created, the FilePolicy job will run as expected. The IndexUpdate job can be run several times daily (for example. every six hours) to keep the index current and prevent the snapshots it uses from growing large.

User configurable file pool policies govern where data is placed, accessed, and protected, accessed, and how it moves among the node pools and tiers. As such, these policies can be used to manage three fundamental properties of data storage:

Property	Description
Location	The physical tier or node pool in which a file lives.
Performance	A file’s performance profile, or I/O optimization setting, which includes sequential, concurrent, or random access. Plus SmartCache write caching
Protection	The protection level of a file, and whether it’s FEC parity-protected or mirrored.

For example, a file pool policy may dictate that anything written to path /ifs/foo goes to the H-Series nodes in node pool 1, then moves to the A-Series nodes in node pool 3 when older than 30 days. The file system itself is doing the work, so there are no transparency or data access risks to worry about.

Also, to simplify management, there are defaults in place for node pool and file pool settings which handle basic data placement, movement, protection and performance. There are several generic template policies, too, which can be customized, cloned, or used as-is

Data movement is parallelized, with the resources of multiple nodes combining for efficient job completion.  While a SmartPools job is running and tiering is in progress, all data is completely available to users and applications.

The performance of node pools can also be governed with SmartPools SSD ‘Strategies’, which can be configured for read caching or metadata storage. Plus the overall system performance impact can be tuned to suit the peaks and lulls of an environment’s workload, by scheduling the SmartPools job to run during off-peak hours.

SmartPools is the OneFS tiering engine, and it enables multiple levels of performance, protection, and storage density to co-exist within a PowerScale cluster. SmartPools allows a cluster admin to define the value of a cluster’s data, and automatically align it with the appropriate price/performance tier over time. Data movement is seamless, and with file-level granularity and control via automated policies, you can easily tune performance and layout, storage tier alignment, and protection settings – with minimal impact to a cluster’s end-users. But first, we’ll run through its taxonomy.

At its core, SmartPools is logically separated into two areas: storage pools and file pools.

Heterogeneous PowerScale clusters can be built with a wide variety of node styles and capacities, in order to meet the needs of a varied data set and wide spectrum of workloads. These node styles fall loosely into three main categories or tiers.

• F-series, all-flash nodes, typically for high performance, low latency workloads
• H-series hybrid nodes, containing a mixture of SSD and hard drives, great for concurrency and streaming workloads.
• A-series active archive nodes, capacity optimized and using large SATA drives.

Storage pools in OneFS provide the ability to define hardware tiers within a single cluster, allowing file layout to be aligned with specific sets of nodes by configuring storage pool policies.

The notion of Storage pools is an abstraction that includes disk pools, node pools, and tiers.

Disk pools are the smallest unit within the storage pools hierarchy. OneFS provisioning works on the premise of dividing the hard drives and SSDs in similar node types into sets, with each pool representing a separate failure domain.

These disk pools are typically protected by default at +2d:1n (or the ability to withstand two disk or one entire node failure) and span a neighborhood from three to forty standalone F-series nodes, or a neighborhood of four to twenty chassis-based H and A series nodes – where each chassis contains four compute modules (one per node), and five drive containers, or ‘sleds’, per node.

Each drive belongs to one disk pool and data protection stripes or mirrors typically don’t extend across pools. Disk pools are managed by OneFS and are generally not user configurable.

Node pools are groups of disk pools, spread across similar storage nodes. Multiple node pools of differing types can coexist in a single, heterogeneous cluster, and this is the lowest level of pool that general SmartPools configuration targets. Say, for example: one node pool of all-flash F-Series nodes for HPC, one node pool of H-Series nodes, for home directories and file shares, and one node pool of A-series nodes, for archive data.

This allows OneFS to present a single storage resource pool, comprising multiple flash and spinning drive media types – NVMe, high speed SAS, large capacity SATA – providing a range of different performance, protection, and capacity characteristics. This heterogeneous storage pools in turn can support a diverse range of applications and workloads with a single, unified namespace and point of management.  It also enables the mixing of older and newer hardware, allowing for simple investment protection even across product generations, and seamless hardware refreshes.

Each node pool only contains disk pools from the same type of storage nodes, and a disk pool may belong to exactly one node pool. For example, all-flash F-series nodes would be in one node pool, whereas A-series nodes with high capacity SATA drives would be in another. Today, a minimum of 4 nodes, or one chassis, are required per node pool for Gen6 modular chassis-based hardware, or three PowerScale F-series nodes per node pool.

Nodes are not associated with each other, or provisioned, until at least three nodes from the same compatibility class are assigned in a node pool. If nodes are removed from a pool, that node pool becomes under-provisioned. In this situation, if two like-nodes remain, they are still writable. If only one remains, it is automatically set to read-only.

Once node pools are created, they can be easily modified to adapt to changing requirements.  Individual nodes can be reassigned from one node pool to another, if necessary.  Node pool associations can also be discarded, releasing member nodes so they can be added to new or existing pools. Node pools can also be renamed at any time without changing any other settings in the node pool configuration.

When new nodes are added to a cluster, they’re automatically allocated to a node pool, and then subdivided into disk pools without any additional configuration steps – and they inherit the SmartPools configuration properties of that node pool. This means the configuration of a pool’s data protection, layout ,and cache settings only needs to be done once, at the time the node pool is first created. Automatic allocation is determined by the shared attributes of the new nodes with the closest matching node pool. If the new node is not a close match to the nodes of any existing pool, it remains un-provisioned until the minimum node pool membership for like-nodes is met.

When a new node pool is created, and nodes are added, SmartPools associates those nodes with a pool ID. This ID is also used in file pool policies and file attributes to dictate file placement within a specific disk pool.

By default, a file which is not covered by a specific file pool policy will go to the configured ‘default’ node pool, identified during set up.  If no default is specified, SmartPools will typically write that data to the pool with the most available capacity.

Tiers are groups of node pools combined into a logical superset to optimize data storage, typically according to OneFS platform type.

For example, similar ‘archive’ node pools are often consolidated into a single tier, which could incorporate different styles of archive node pools into a single, logical container. For example, PowerScale A300s with 12TB SATA drives and PowerScale A3000s with 16TB SATA drives logically combined into a single active archive tier. This is a significant benefit to customers who consistently purchase the highest capacity nodes available, to consolidate a variety of node styles within a single tier and manage them as one logical group.

Note, however, that a storage efficiency cost may be incurred if the node pools in a tier are too small. For example, in a six node cluster with two separate three-node pools (different drive sizes), each pool has a 33% protection overhead. This is compared to a six node cluster with single six-node pool (same drive size), protection overhead drops to 16% (at the default +2d:1n protection).

SmartPools users frequently deploy 2 to 4 tiers, with the fastest tier typically containing all-flash nodes for the most performance demanding portions of a workflow, and the lowest, capacity-biased tier comprising high capacity SATA drive nodes.

SmartPools allows nodes of any type supported by the particular OneFS version, to be combined within the same cluster. The like-nodes are provisioned into different node pools according to their physical attributes:

These node compatibility classes are fairly stringent. This is in order to avoid disproportionate amounts of work being directed towards a subset of cluster resources, which could result in bullying of the lower powered nodes.

However, cluster administrators can safely target specific data to broader classes of storage by creating tiers. For example, if a cluster includes two different varieties of H nodes, such as H700s and H7000s, these will automatically be provisioned into two different node pools. These two node pools can be logically combined into a tier, and file placement targeted to it, resulting in automatic balancing across the node pools.

SmartPools separates hardware by node type and creates a separate node pool for each distinct hardware variant. To reside in the same node pool, nodes must have a set of core attributes in common, and node compatibilities can be defined to allow nodes with the same drive types, quantities and capacities and compatible RAM configurations, to be provisioned into the same pools.

That said, due to significant architectural differences, there are no node compatibilities between the chassis-based all-flash F800 or F810s, and the self-contained all-flash nodes like the F600 or F900.

OneFS also contains an SSD compatibility option, which allows nodes with dissimilar flash capacity to be provisioned to a single node pool. When creating this SSD compatibility, OneFS automatically checks that the two pools to be merged have the same number of SSDs, tier, requested protection, and the same SSD strategy or L3 cache setting.

If a node pool fills up, writes to that pool will automatically spill over to the next pool.  This default behavior ensures that work can continue, even if one type of capacity is full.  There are some circumstances in which spillover is undesirable, for example when different business units within an organization purchase separate pools, or data location has security or protection implications.  In these circumstances, spillover can simply be disabled.  Disabling spillover ensures a file exists in one pool and will not move to another.

From a data protection and layout efficiency point of view, SmartPools subdivides large numbers of like nodes into smaller, more efficiently protected disk pools – automatically calculating and grouping the cluster into pools of disks, that are optimized for both Mean Time to Data Loss (MTTDL) and efficient space utilization. This means that protection level decisions are not left to the cluster admin, unless desired.

With Automatic Provisioning, every set of equivalent node hardware is automatically split up into disk pools, node pools and neighborhoods. These pools are protected by default against up to two drive or one node failure per disk pool. By subdividing a node’s disks into multiple, separately protected disk pools, nodes are significantly more resilient to multiple disk failures.

If the automatically provisioned node pools that OneFS creates are not appropriate for an environment, they can be manually reconfigured. This is done by creating a manual node pool and moving nodes from an existing node pool to the newly created one. However, the strong recommendation is to use the default, automatically provisioned node pools. Manually assigned pools may not provide the same level of performance and storage efficiency as automatically assigned pools.

Unlike hardware RAID, OneFS has no requirement for dedicated hot spare drives. Instead, it simply borrows from the available free space in the file system in order to recover from failures; this technique is called virtual hot spare, or VHS.

SmartPools Virtual Hot Spare helps ensure that node pools maintain enough free space to successfully re-protect data in the event of drive failure. Though configured globally, VHS actually operates at the disk pool level so that nodes with different size drives reserve the appropriate VHS space. This helps ensure that, while data may move from one disk pool to another during repair, it remains on the same class of storage.

VHS reservations are cluster wide and configurable as either a percentage of total storage, up to 20%, or from 1 to 4 virtual drives. This reservation works by allocating a fraction of the node pool’s VHS space in each of its constituent disk pools.

Keep in mind that reservations for virtual hot sparing will affect spillover – if, for example, VHS is configured to reserve 10% of a pool’s capacity, spillover will occur at 90% full.

Introduced in OneFS 9.3, SMB drain support further enhances OneFS non-disruptive upgrades, by allowing for the safe disconnection of SMB clients. In an ideal world, OneFS would be able to seamlessly migrate all SMB clients transparently to non-rebooting nodes. Windows continuous availability (CA) does this natively, but this is not always a viable option given the client SMB3 support requirements, the performance implications of CA, etc.

Because SMB clients may be caching data through the use of oplocks or leases, it is important to ensure that this caching is stopped prior to disconnecting a client. OneFS SMB drain support ensures that, in non-CA cases, an SMB client is able to flush its cache before being disconnected, and, in conjunction with SmartConnect, enables safe migration of SMB clients to non-rebooting nodes in a cluster.

The following diagram illustrates the basic interaction of the SMB server and SmartConnect with the drain service:

Both SmartConnect and SMB detect when the drain service is running on the local node through the OneFS group management protocol (GMP). When the drain service is active, SmartConnect will no longer include the draining node’s IP address in DNS query responses. Then SMB starts the process of disconnecting clients.

GMP indicates whether the drain service is running on a local node, and, if so, SMB will no longer grant new oplocks or leases. So when a new oplock or lease is requested, the server responds indicating a conflict, which prevents the granting of the lock or lease. SMB then starts the process of breaking existing oplocks and leases, by emulating conflicting access. So, in the case of the lease,  OneFS will send a break response to the client, and, depending on the type of lease, will either wait for an acknowledgement of the break or break the lease immediately. The OneFS lwio server continually scans for sessions which have no oplocks or leases, and these sessions can then be drained down and disconnected. Note that an SMB clients with oplocks and leases disabled will automatically be a candidate for disconnection, since no sessions or locks will be detected during the scan. When a session is disconnected, the drain service notes the time of the disconnection and the client’s GUID (in the case of SMB 2 or 3) or its IP address (if SMB1). This information is used to track any reconnecting clients.

Once a Windows client has been disconnected, it typically sends a DNS request and receives  a response with an IP address of a non-draining node to connect to. However, not all SMB clients do this. For example, Linux and MacOS SMB clients will often perform a small number of attempts to reconnect to the previous IP address instead (either due to caching or stubborn client behavior), before declaring a network error. This is obviously undesirable since it results in a user-visible event. So OneFS cannot immediately disconnect reconnecting clients. Instead, the client is allowed to reconnect, but the lwio server starts delaying the responses to ‘session setup’ and ‘tree connect’ requests by 8 seconds by default. So this limits what the client can do after reconnect, and the goal is two persuade it to send a DNS request instead and connect to a non-rebooting node. The responses to negotiate request are not delayed because most clients will automatically consider the lack of a response as a network error and will not retry. If the node happens to reboot before the negotiation response is sent, the client will likely report an error to the user, so OneFS does not delay the response to minimize this possibility. The server still will not allow oplocks and leases to be granted, and will eventually disconnect the client again, after a default 20 seconds since the last time the client was disconnected.

No configuration is required for this SMB drain functionality in OneFS 9.3 and later, and, as such, there are no CLI commands to control it, etc. The drain service is started on the local node and the service is going to go through the process of safely disconnecting the clients. However, there is a OneFS registry parameter which, if necessary, can be used to modify or override the SMB drain behavior via the isi_gconfig CLI command.

For example, to disable SMB draining:

# isi_gconfig registry.Services.lwio.Parameters.Drivers.srv.EnableSessionDraining=0

registry.Services.lwio.Parameters.Drivers.srv.EnableSessionDraining (uint32) = 0

The three configurable values are:

Parameter	Default Value	Description
EnableSessionDraining	Default is ‘enabled’.	Global SMB draining on/off switch.
DrainDisconnectTimeout	Default is 20 seconds.	Controls the minimum time between disconnecting and reconnecting clients.
DrainResponseDelay	Default is 8 seconds.	Controls the delay period for responses to ‘session setup’ and ‘tree connect’ requests.

Be aware that, unlike DrainDisconnectTimeout which is in seconds, the DrainResponseDelay parameter is expressed in milliseconds (ms):

# isi_gconfig registry.Services.lwio.Parameters.Drivers.srv.DrainResponseDelay

registry.Services.lwio.Parameters.Drivers.srv.DrainResponseDelay (uint32) = 8000

SMB safe disconnect works in concert with OneFS’ drain-based upgrade, which was introduced in OneFS 9.2. Drain-based upgrade provides a mechanism by which nodes are prevented from rebooting or restarting protocol services until all SMB clients have disconnected from the node. A single SMB client that does not disconnect can cause the upgrade to be delayed indefinitely, so the cluster administrator is provided with options to reboot the node despite persisting clients.

As a truly non-disruptive upgrade process, drain-based upgrade can be potentially slower, since it is dependent upon client disconnections. The core OneFS protocols are handled as follows:

Protocol	Action
SMB	Wait for clients to drain and disconnect before rebooting node
SMB3-CA	Witness, drain service → immediate migration → faster upgrade
NFS, HDFS, HTTP, S3	Assumed resilient to rebooting nodes

Drain-based upgrades can be configured and managed via the OneFS WebUI, CLI, and RESTful platform API, and the supported operations include:

• OneFS upgrades
• Firmware upgrades
• Cluster reboots
• Combined upgrades (OneFS and Firmware)

Drain-based upgrade is predicated upon the parallel upgrade workflow, which offers accelerated upgrades for large clusters by working across OneFS neighborhoods, or fault domains. By concurrently upgrading a node per neighborhood, the more node neighborhoods there are within a cluster the more parallel activity can occur.

Imagine a PowerScale H700 cluster with five chassis split into two neighborhoods, each containing ten nodes:

Once the drain-based upgrade is started, a maximum of one node from each neighborhood will get the reservation, which allows the nodes to upgrade simultaneously. OneFS will not reboot these nodes until the number of SMB clients is “0”. Say nodes 12 and 17 get the reservation for upgrading at the same time. However, there is one SMB connection to node 17 and two SMB connections to node 12. Neither of these nodes will be able to reboot until their SMB connection count gets to “0”. At this point, there are three options available:

Drain Action	Description
Wait	Wait until the SMB connection count reaches “0” or it hits the drain timeout value. The drain timeout value is a configurable parameter for each upgrade process. It is the maximum waiting period. If drain timeout is set to “0”, it means wait forever.
Delay drain	Add the node into the delay list to delay client draining. The upgrade process will continue on another node in this neighborhood. After all the non-delayed nodes are upgraded, OneFS will rewind to the node in the delay list.
Skip drain	Stop waiting for clients to migrate away from the draining node and reboot immediately.

The ‘isi upgrade cluster drain’ CLI command syntax can be used to manage client draining per-node. For example, to configure node 1 in the cluster to delay draining:

# isi upgrade cluster drain delay add 1

The node(s) will delay draining active SMB client connections (until all nodes in the same neighborhood have finished draining). Are you sure? (yes/[no]): yes

# isi upgrade cluster drain delay list

LNN

----

1

The following CLI syntax can be used to confirm whether there are any active SMB connections. In this case, node 1 has one connected Windows client:

# isi statistics query current --keys=node.clientstats.connected.smb

Node  node.clientstats.connected.smb

-------------------------------------

    1                               1

-------------------------------------

The ‘isi upgrade’ CLI command syntax can be used to perform the drain-based upgrade, and now includes flags for configuring drain-timeout and alert-timeout. In this example setting each to value 60 minutes and 45 minutes respectively. As such, if there is still an SMB connection after 45 minutes, a CELOG alert will be triggered to notify the cluster administrator. And after an hour, any remaining SMB connections will be dropped, and the node upgrade reboot will continue.

# isi upgrade start --parallel --skip-optional --install-image-path=/ifs /data/<installation-file-name> --drain-timeout=60m --alert-timeout=45m

From the OneFS WebUI, the same can be achieved by navigating to Upgrade under Cluster management.

In the previous articles in this series, we’ve looked at the fundamentals of a cluster’s network infrastructure:

The complete cluster architecture – software, hardware, and network – all cooperate to provide a distributed single file system that can scale dynamically as workloads and capacity and/or throughput needs change in a scale-out environment.

OneFS SmartConnect provides the load balancing services that work at the front-end Ethernet layer to evenly distribute client connections across the cluster. SmartConnect supports dynamic NFS failover and failback for Linux and UNIX clients and SMB3 continuous availability for Windows clients. This ensures that when a node failure occurs, or preventative maintenance is performed, all in-flight reads and writes are handed off to another node in the cluster to finish its operation without any user or application interruption.

During failover, clients are evenly redistributed across all remaining nodes in the cluster, ensuring minimal performance impact. If a node is brought down for any reason, including a failure, the virtual IP addresses on that node is seamlessly migrated to another node in the cluster.

When the offline node is brought back online, SmartConnect automatically rebalances the NFS and SMB3 clients across the entire cluster to ensure maximum storage and performance utilization. For periodic system maintenance and software updates, this functionality allows for per-node rolling upgrades affording full-availability throughout the duration of the maintenance window.

The OneFS SmartConnect module itself can be run in two modes – with or without a license:

SmartConnect Attribute	SmartConnect Basic (unlicensed)	SmartConnect Advanced (Licensed)
Connection Balancing	Round-robin only.	Round-robin, CPU utilization, connection counting, and throughput balancing.
Address Allocation	Static IP allocation only.	Static and dynamic IP address allocation, up to a maximum of six SmartConnect Service IP addresses per subnet.
Address Failover	No IP address failover policy.	Supports defining a failover policy for the IP address pool.
Address Rebalance	No IP address rebalance policy.	Supports defining a rebalance policy for the IP address pool.
Per-pool Addresses	Up to two IP address pools per external network subnet	Supports multiple IP address pools per external subnet to enable multiple DNS zones within a single subnet.

The SmartPools static vs dynamic address allocation method indicates whether the IP addresses in the pool can move back and forth between nodes when a node goes down. As such, a static IP pool displays the following characteristics:

• Each interface in the pool gets exactly one IP (assuming there are at least as many IPs as interfaces in the pool).
• If there are more IPs in the pool than interfaces, the additional IPs will not be allocated to any interface.
• IPs do not move from one interface to another.
• If an interface goes down, then the IP also goes down.

Conversely, in a dynamic IP pool:

• Each of the IPs in the pool is allocated to an interface in the pool.
• When an interface goes down in the pool, the IPs on that interface automatically move to another interface in the pool (preferring interfaces in the pool that are on the same node as the downed interface).
• When a node is transitions to an ‘unhealthy’ state, the IPs on that node automatically move to another node in the pool.
• When a node transitions back to a ‘healthy’ state, IPs will automatically move back to that node, assuming the rebalance policy is set to ‘auto’ and there are enough IPs available.

By default, OneFS SmartConnect balances connections among nodes using a round-robin policy and a separate IP pool for each subnet. A SmartConnect license adds advanced balancing policies to evenly distribute CPU usage, client connections, or throughput. It also lets you define IP address pools to support multiple DNS zones in a subnet.

Load-balancing Policy	General	Few Clients with High Usage	Many Persistent NFS & SMB Connections	Many Ephemeral Connections (HTTP, FTP)	NFS Automount of UNC Paths are Used
Round Robin (Default)					
Connection Count					
CPU Usage
Network Throughput

Connection policies other than round robin are sampled every 10 seconds. The CPU policy is sampled every 5 seconds. If multiple requests are received during the same sampling interval, SmartConnect will attempt to balance these connections by estimating or measuring the additional load.

A ‘round robin’ load balancing strategy is generally a safe bet for both client connection balancing and IP failover.

Under the hood, SmartConnect acts as DNS delegation server, responding to requests and returning IP addresses for the appropriate SmartConnect zone(s). The general transactional flow is as follows:

During a cluster ‘split’ or ‘merge’ group change the SmartConnect service will not respond to DNS inquiries. This is seldom as group changes typically take around 30 seconds. However, the time taken for a group change to complete can vary due to the load on the cluster at the time of the change. Any time a node is added, removed, or rebooted in a cluster there will be two group changes that cause SmartConnect to be impacted, one from down/split and one from up/merge.

For large clusters, if group changes are adversely impacting SmartConnect’s load-balancing performance, the core site DNS servers can be configured to use a Round Robin configuration instead of redirecting DNS requests to SmartConnect

SmartConnect supports IP failover to provide continuous access to data when hardware or a network path fails. Dynamic failover is recommended for high availability workloads on SmartConnect subnets that handle traffic from NFS clients.

For optimal network performance, avoid mixing interface types (100/40/25/10 GbE) in the same SmartConnect Pool and/or mixing node types with different performance profiles, such as H700 and A300 interfaces, for example. In general, the ‘round-robin’ SmartConnect Client Connection Balancing and IP-failover policies provide the most consistent results.

To evenly distribute connections and optimize performance, the recommendation is to size SmartConnect for the expected number of connections and for the anticipated overall throughput likely to be generated. The sizing factors for a pool include the total number of concurrently active client connections, the anticipated aggregate throughput for the pool, and he minimum performance and throughput requirements in case an interface fails.

Since OneFS is a single volume, fully distributed file system, a client can access all the files and associated metadata that are stored on the cluster, regardless of the type of node a client connects to or the node pool on which the data resides. For example, data stored for performance reasons on a pool of F-Series all-flash nodes can be mounted and accessed by connecting to an A-Series node in the same cluster. The different types of PowerScale nodes, however, deliver different levels of performance.

To avoid unnecessary network latency under most circumstances, the recommendation is to configure SmartConnect subnets such that client connections are to the same physical pool of nodes on which the data resides. In other words, if a workload’s data lives on a pool of F600 nodes for performance reasons, the clients that work with that data should mount the cluster through a pool that includes the same F600 nodes that host the data.

Keep in mind the following networking and name server considerations:

• Minimize disruption by suspending nodes in preparation for planned maintenance and resuming them after maintenance is complete
• Leverage the groupnet feature to enhance multi-tenancy and DNS delegation, where desirable.
• Ensure traffic flows through the right interface by tracing routes. Leverage OneFS Source-Based Routing (SBR) feature to keep traffic on desired paths.

If firewalling or filtering is deployed within the network, ensure that the appropriate ports are open. For example, open both UDP port 53 and TCP port 53 for the DNS service.

The client never sends a DNS request directly to the cluster. Instead, the site nameservers handle DNS requests from clients and route the requests appropriately.

In order to successfully distribute IP addresses, the OneFS SmartConnect DNS delegation server answers DNS queries with a time-to-live (TTL) of 0 so that the answer is not cached. Certain DNS servers (particularly Windows DNS Servers) will fix the value to one second. If you have many clients requesting an address within the same second, this will cause all of them to receive the same address. If you encounter this problem, you may need to use a different DNS server, such as BIND.

Certain clients perform DNS caching and might not connect to the node with the lowest load if they make multiple connections within the lifetime of the cached address. Recommend turning off client DNS caching, where possible. To handle client requests properly, SmartConnect requires that clients use the latest DNS entries.

The site DNS servers must be able to communicate with the node that is currently hosting the SmartConnect service. This is the node with the lowest logical node number (LNN) with an active interface in the subnet that contains the SSIP address.

As we’ve seen in prior articles in this series, OneFS and the PowerScale platforms support a variety of Ethernet speeds, cable and connector styles, and network interface counts, depending on the node type selected. However, unlike the back-end network, Dell does not specify particular front-end switch models, allowing PowerScale clusters to seamlessly integrate into the data link layer (layer 2) of an organization’s existing Ethernet IP network infrastructure. For example:

A layer 2 looped topology as above extends VLANs between the distribution/aggregation switches, with spanning tree protocol (STP) preventing network loops by shutting down redundant paths. The access layer uplinks may be used to load balance VLANs. This distributed architecture allows the cluster’s external network to connect to multiple access switches, affording each node similar levels of availability, performance, and management properties.

Link aggregation can be used to combine multiple Ethernet interfaces into a single link-layer interface, and is implemented between a single switch and  PowerScale node, where transparent failover or switch port redundancy is required. Link aggregation assumes all links are full duplex, point to point, and at the same data rate, providing graceful recovery from link failures. If a link fails, traffic is automatically sent to the next available link without disruption.

Quality of service (QoS) can be implemented through differentiated services code point (DSCP), by specifying a value in the packet header that maps to an ‘effort level’ for traffic. Since OneFS does not provide an option for tagging packets with a specified DSCP marking, the recommended practice is to configure the first hop ports to insert DSCP values on the access switches connected to the PowerScale nodes. OneFS does retain headers for packets that already have a specified DSCP value, however.

When designing a cluster, the recommendation is that each node have at least one front-end interface configured, preferably in at least one static SmartConnect zone. Although a cluster can be run in a ‘not all nodes on the network’ (NANON) configuration, where feasible, the recommendation is to connect all nodes to the front-end network(s). Additionally, cluster services such as SNMP, ESRS, ICAP, and auth providers (AD, LDAP, NIS, etc) prefer each node to have an address that can reach the external servers.

In contrast with scale-up NAS platforms that use separate network interfaces for out-of-band management and configuration, OneFS traditionally performs all cluster network management in-band. However, PowerScale nodes typically contain a dedicated 1Gb Ethernet port that can be configured for use as a management network via ICMP or iDRAC, simplifying administration of a large cluster. OneFS also supports using a node’s serial port as an RS232 out-of-band management interface, and this practice is highly recommended for large clusters. Serial connectivity can provide reliable BIOS-level command line access for on-site or remote service staff to perform maintenance, troubleshooting and installation operations.

SmartConnect provides a configurable allocation method for each IP address pool:

Allocation Method	Attributes
Static	• One IP per interface is assigned, will likely require fewer IP’s to meet minimum requirements • No Failover of IP’s to other interfaces
Dynamic	• Multiple IP per interface is assigned, will require more IP’s to meet minimum requirements • Failover of IP’s to other interfaces, Failback policies are needed

The default ‘static’ allocation assigns a single persistent IP address to each interface selected in the pool, leaving additional pool IP addresses  unassigned if the number of addresses exceeds the total interfaces.

The lowest IP address of the pool is assigned to the lowest Logical Node Number (LNN) from the selected interfaces, subsequently for the second-lowest IP address and LNN, etc. If a node or interface becomes unavailable, this IP address does not move to another node or interface. Also, when the node or interface becomes unavailable, it is removed from the SmartConnect zone, and new connections will not be assigned to the node. Once the node is available again, SmartConnect will automatically add it back into the zone and assign new connections.

By contrast, ‘dynamic’ allocation divides all available IP addresses in the pool across all selected interfaces, and OneFS attempts to assign the IP addresses as evenly as possible. However, if the interface-to-IP address ratio is not an integer value, a single interface might have more IP addresses than another. As such, wherever possible, ensure that all the interfaces have the same number of IP addresses.

In concert with dynamic allocation, dynamic failover provides high availability by transparently migrating IP addresses to another node when an interface is not available. If a node becomes unavailable, all the IP addresses it was hosting are reallocated across the new set of available nodes in accordance with the configured failover load-balancing policy. The default IP address failover policy is round robin, which evenly distributes IP addresses from the unavailable node across available nodes. Because the IP address remains consistent, irrespective of which node it resides on, failover to the client is transparent, so high availability is seamless.

The other available IP address failover policies are the same as the initial client connection balancing policies, that is, connection count, throughput, or CPU usage. In most scenarios, round robin is not only the best option but also the most common. However, the other failover policies are available for specific workflows.

The decision on whether to implement dynamic failover is highly dependent on the protocol(s) being used, general workflow attributes, and any high-availability design requirements:

Protocol	State	Suggested Allocation Strategy
NFSv3	Stateless	Dynamic
NFSv4	Stateful	Dynamic or Static, depending on mount daemon, OneFS version, and Kerberos.
SMB	Stateful	Dynamic or Static
SMB Multi-channel	Stateful	Dynamic or Static
S3	Stateless	Dynamic or Static
HDFS	Stateful	Dynamic or Static. HDFS uses separate name-node and data-node connections. Allocation strategy depends on need for data locality and/or multi-protocol. Ie: HDFS + NFSv3 : Dynamic Pool HDFS + SMB : Static Pool
HTTP	Stateless	Static
FTP	Stateful	Static
SyncIQ	Stateful	Static required

Assigning each workload or data store to a unique IP address enables OneFS SmartConnect to move each workload to one of the other interfaces, minimizing the additional work that a remaining node in the SmartConnect pool must absorb and ensuring that the workload is evenly distributed across all the other nodes in the pool.

Static IP pools require one IP address for each logical interface within the pool. Since each node provides two interfaces for external networking, if link aggregation is not configured, this would require 2*N IP addresses for a static pool.

Determining the number of IP addresses within a dynamic allocation pool varies depending on the workflow, node count, and the estimated number of clients that would be in a failover event. While dynamic pools need, at a minimum, the number of IP addresses to match a pool’s node count, the ‘N * (N – 1)’ formula can often prove useful for calculating the required number of IP addresses for smaller pools. In this equation, N is the number of nodes that will participate in the pool

For example,  a SmartConnect pool with four-node interfaces, using the ‘N * (N – 1)’ model will result in three unique IP addresses being allocated to each node. A failure on one node interface will cause each of that interface’s three IP addresses to fail over to a different node in the pool. This ensures that each of the three active interfaces remaining in the pool receives one IP address from the failed node interface. If client connections to that node are evenly balanced across its three IP addresses, SmartConnect will evenly distribute the workloads to the remaining pool members. For larger clusters, this formula may not be feasible due to the sheer number of IP addresses required.

Enabling jumbo frames (Maximum Transmission Unit set to 9000 bytes) typically yields improved throughput performance with slightly reduced CPU usage than standard frames, where the MTU is set to 1500 bytes. For example, with 40 Gb Ethernet connections, jumbo frames provide about five percent better throughput and about one percent less CPU usage.

OneFS provides the ability to optimize storage performance by designating zones to support specific workloads or subsets of clients. Different network traffic types can be segregated on separate subnets using SmartConnect pools.

For large clusters, partitioning the cluster’s networking resources and allocating bandwidth to each workload can help minimize the likelihood that heavy traffic from one workload will affect network throughput for another. This is particularly true for SyncIQ replication and NDMP backup traffic, which can frequently benefit from its own set of interfaces, separate from user and client IO load.

The ‘groupnet’ networking object is part of OneFS’ support for multi-tenancy. Groupnets sit above subnets and pools and allow separate Access Zones to contain distinct DNS settings.

The management and data network(s) can then be incorporated into different Access Zones, each with their own DNS, directory access services, and routing, as appropriate.

A key decision for performance, particularly in a large cluster environment, is the type and quantity of nodes deployed. Heterogeneous clusters can be architected with a wide variety of node styles and capacities, in order to meet the needs of a varied data set and wide spectrum of workloads. These node styles encompass several hardware generations, and fall loosely into three main categories or tiers. While heterogeneous clusters can easily include many hardware classes and configurations, the best practice of simplicity for building clusters holds true here too.

Consider the physical cluster layout and environmental factors, particularly when designing and planning a large cluster installation. These factors include:

• Redundant power supply
• Airflow and cooling
• Rackspace requirements
• Floor tile weight constraints
• Networking Requirements
• Cabling distance Limitations

The following table details the physical dimensions, weight, power draw, and thermal properties for the range of PowerScale F-series all-flash nodes:

Model	Tier	Height	Width	Depth	RU	Weight	MaxWatts	Watts	Max BTU	Normal BTU
F900	All-flash NVMe performance	2U (2×1.75IN)	17.8 IN / 45 cm	31.8 IN / 85.9 cm	2RU	73 lbs	1297	859	4425	2931
F600	All-flash NVMe Performance	1U (1.75IN)	17.8 IN / 45 cm	31.8 IN / 85.9 cm	1RU	43 lbs	467	718	2450	1594
F200	All-flash Performance	1U (1.75IN)	17.8 IN / 45 cm	31.8 IN / 85.9 cm	1RU	47 lbs	395	239	1346	816

 

Note that the table above represents individual nodes. A minimum of three of each node style are required for a node pool.

Similarly, the following table details the physical dimensions, weight, power draw, and thermal properties for the range of PowerScale chassis-based platforms:

Model	Tier	Height	Width	Depth	RU	Weight	MaxWatts	Watts	Max BTU	Normal BTU
F800/810	All-flash performance	4U (4×1.75IN)	17.6 IN / 45 cm	35 IN / 88.9 cm	4RU	169 lbs (77 kg)	1764	1300	6019	4436
H700	Hybrid/Utility	4U (4×1.75IN)	17.6 IN / 45 cm	35 IN / 88.9 cm	4RU	261lbs (100 kg)	1920	1528	6551	5214
H7000	Hybrid/Utility	4U (4×1.75IN)	17.6 IN / 45 cm	39 IN / 99.06 cm	4RU	312 lbs (129 kg)	2080	1688	7087	5760
H600	Hybrid/Utility	4U (4×1.75IN)	17.6 IN / 45 cm	35 IN / 88.9 cm	4RU	213 lbs (97 kg)	1990	1704	6790	5816
H5600	Hybrid/Utility	4U (4×1.75IN)	17.6 IN / 45 cm	39 IN / 99.06 cm	4RU	285 lbs (129 kg)	1906	1312	6504	4476
H500	Hybrid/Utility	4U (4×1.75IN)	17.6 IN / 45 cm	35 IN / 88.9 cm	4RU	248 lbs (112 kg)	1906	1312	6504	4476
H400	Hybrid/Utility	4U (4×1.75IN)	17.6 IN / 45 cm	35 IN / 88.9 cm	4RU	242 lbs (110 kg)	1558	1112	5316	3788
A300	Archive	4U (4×1.75IN)	17.6 IN / 45 cm	35 IN / 88.9 cm	4RU	252 lbs (100 kg)	1460	1070	4982	3651
A3000	Archive	4U (4×1.75IN)	17.6 IN / 45 cm	39 IN / 99.06 cm	4RU	303 lbs (129 kg)	1620	1230	5528	4197
A200	Archive	4U (4×1.75IN)	17.6 IN / 45 cm	35 IN / 88.9 cm	4RU	219 lbs (100 kg)	1460	1052	4982	3584
A2000	Archive	4U (4×1.75IN)	17.6 IN / 45 cm	39 IN / 99.06 cm	4RU	285 lbs (129 kg)	1520	1110	5186	3788

 

Note that the table above represents 4RU chassis, each of which contains four PowerScale platform nodes (the minimum node pool size).

Below are the locations of both the front end (ext-1 & ext-2) and back-end (int-1 & int-2) network interfaces on the PowerScale stand-alone F-series and chassis-based nodes:

A PowerScale cluster’s backend network is analogous to a distributed systems bus. Each node has two backend interfaces for redundancy that run in an active/passive configuration (int-1 and int-2 above). The primary interface is connected to the primary switch, and the secondary interface to a separate switch.

For nodes using 40/100 Gb or 25/10 Gb Ethernet or Infiniband connected with multimode fiber, the maximum cable length is 150 meters. This allows a cluster to span multiple rack rows, floors, and even buildings, if necessary. While this can solve floor space challenges, in order to perform any physical administration activity on nodes you must know where the equipment is located.

The table below shows the various PowerScale node types and their respective backend network support. While Ethernet is the preferred medium – particularly for large PowerScale clusters –Infiniband is also supported for compatibility with legacy Isilon clusters.

Node Models	Details
F200, F600, F900	F200: nodes support a 10 GbE or 25 GbE connection to the access switch using the same NIC. A breakout cable can connect up to four nodes to a single switch port. F600: nodes support a 40 GbE or 100 GbE connection to the access switch using the same NIC. F900: nodes support a 40 GbE or 100 GbE connection to the access switch using the same NIC.
H700, H7000, A300, A3000	Supports 40 GbE or 100 GbE connection to the access switch using the same NIC. OR Supports 25 GbE or 10 GbE connection to the leaf using the same NIC. A breakout cable can connect a 40 GbE switch port to four 10 GbE nodes or a 100 GbE switch port to four 25 GbE nodes.
F810, F800, H600, H500, H5600	Performance nodes support a 40 GbE connection to the access switch.
A200, A2000, H400	Archive nodes support a 10GbE connection to the access switch using a breakout cable. A breakout cable can connect a 40 GbE switch port to four 10 GbE nodes or a 100 GbE switch port to four 10 GbE nodes.

Currently only Dell approved switches are supported for backend Ethernet and IB cluster interconnection. These include:

Switch Model	Port Count	Port Speed	Height (Rack Units)	Role	Notes
Dell S4112	24	10GbE	½	ToR	10 GbE only.
Dell 4148	48	10GbE	1	ToR	10 GbE only.
Dell S5232	32	100GbE	1	Leaf or Spine	Supports 4x10GbE or 4x25GbE breakout cables. Total of 124 10GbE or 25GbE nodes as top-of-rack backend switch. Port 32 does not support breakout.
Dell Z9100	32	100GbE	1	Leaf or Spine	Supports 4x10GbE or 4x25GbE breakout cables. Total of 128 10GbE or 25GbE nodes as top-of-rack backend switch.
Dell Z9264	64	100GbE	2	Leaf or Spine	Supports 4x10GbE or 4x25GbE breakout cables. Total of 128 10GbE or 25GbE nodes as top-of-rack backend switch.
Arista 7304	128	40GbE	8	Enterprise core	40GbE or 10GbE line cards.
Arista 7308	256	40GbE	13	Enterprise/ large cluster	40GbE or 10GbE line cards.
Mellanox Neptune MSX6790	36	QDR	1	IB fabric	32Gb/s quad data rate Infiniband.

Be aware that the use of patch panels is not supported for PowerScale cluster backend connections, regardless of overall cable lengths. All connections must be a single link, single cable directly between the node and backend switch. Also, Ethernet and Infiniband switches must not be reconfigured or used for any traffic beyond a single cluster.

Support for leaf spine backend Ethernet network topologies was first introduced in OneFS 8.2. In a leaf-spine network switch architecture, the PowerScale nodes connect to leaf switches at the access, or leaf, layer of the network. At the next level, the aggregation and core network layers are condensed into a single spine layer. Every leaf switch connects to every spine switch to ensure that all leaf switches are no more than one hop away from one another. For example:

Leaf-to-spine switch connections require even distribution, to ensure the same number of spine connections from each leaf switch. This helps minimize latency and reduces the likelihood of bottlenecks in the back-end network. By design, a leaf spine network architecture is both highly scalable and redundant.

Leaf spine network deployments can have a minimum of two leaf switches and one spine switch. For small to medium clusters in a single rack, the back-end network typically uses two redundant top-of-rack (ToR) switches, rather than implementing a more complex leaf-spine topology.

When it comes to physically installing PowerScale nodes, most utilize a 35 inch depth chassis and will fit in a standard depth data center cabinet. Nodes can be secured to standard storage racks with their sliding rail kits, included in all node packaging and compatible with racks using either 3/8 inch square holes, 9/32 inch round holes, or 10-32 / 12-24 / M5X.8 / M6X1 pre-threaded holes. These supplied rail kit mounting brackets are adjustable in length from 24 inches to 36 inches to accommodate different rack depths. When selecting an enclosure for PowerScale nodes, ensure that the rack supports the minimum and maximum rail kit sizes.

Rack Component	Description
a	Distance between front surface of the rack and the front NEMA rail
b	Distance between NEMA rails, minimum=24in (609.6mm), max=34in (863.6mm)
c	Distance between the rear of the chassis to the rear of the rack, min=2.3in (58.42mm)
d	Distance between inner front of the front door and the NEMA rail, min=2.5in (63.5mm)
e	Distance between the inside of the rear post and the rear vertical edge of the chassis and rails, min=2.5in (63.5mm)
f	Width of the rear rack post
g	19in (486.2mm)+(2e), min=24in (609.6mm)
h	19in (486.2mm) NEMA+(2e)+(2f) Note: Width of the PDU+0.5in (13mm) <=e +f If j=i+c+PDU depth+3in (76.2mm), then h=min 23.6in (600mm) Assuming the PDU is mounted beyond i+c.
i	Chassis depth: Normal chassis=35.80in (909mm) : Deep chassis=40.40in (1026mm) Switch depth (measured from the front NEMA rail): Note: The inner rail is fixed at 36.25in (921mm) Allow up to 6in (155mm) for cable bend radius when routing up to 32 cables to one side of the rack. Select the greater of the installed equipment.
j	Minimum rack depth=i+c
k	Front
l	Rear
m	Front door
n	Rear door
p	Rack post
q	PDU
r	NEMA
s	NEMA 19 inch
t	Rack top view
u	Distance from front NEMA to chassis face: Dell PowerScale deep and normal chassis = 0in

However, the high capacity models such as the F800/810, H7000, H5600, A3000 and A2000 have 40 inch depth chassis and require extended depth cabinets such as the APC 3350 or Dell Titan-HD rack.

Additional room must be provided for opening the FRU service trays at the rear of the nodes and, in the chassis-based 4RU platforms, the disk sleds at the front of the chassis. With the exception of the 2RU F900, the stand-alone PowerScale all-flash nodes are 1RU in height (including the 1RU diskless P100 accelerator and B100 backup accelerator nodes).

Power-wise, each cabinet typically requires between two and six independent single or three-phase power sources. To determine the specific requirements, use the published technical specifications and device rating labels for the devices to calculate the total current draw for each rack.

Specification	North American 3 wire connection (2 L and 1 G)	International 3 wire connection (1 L, 1 N, and 1 G)
Input nominal voltage	200–240 V ac +/- 10% L – L nom	220–240 V ac +/- 10% L – L nom
Frequency	50–60 Hz	50–60 Hz
Circuit breakers	30 A	32 A
Power zones	Two	Two
Power requirements at site (minimum to maximum)	Single-phase: six 30A drops, two per zone Three-phase Delta: two 50A drops, one per zone Three-phase Wye: two 32A drops, one per zone	Single-phase: six 30A drops, two per zone Three-phase Delta: two 50A drops, one per zone Three-phase Wye: two 32A drops, one per zone

Additionally, the recommended environmental conditions to support optimal PowerScale cluster operation are as follows:

Attribute	Details
Temperature	Operate at >=90 percent of the time between 10 degrees celsiuses to 35 degrees celsius degrees celsius, and <=10 percent of the time between 5 degrees celsiuses to 40 degrees celsiuses.
Humidity	40 to 55 percent relative humidity
Weight	A fully configured cabinet must sit on at least two floor tiles, and can weigh approximately 1588 kilograms (3500 pounds).
Altitude	0 meters to 2439 meters (0 to 8,000 ft) above sea level operating altitude.

Weight is a critical factor to keep in mind, particularly with the chassis-based nodes. Individual 4RU chassis can weigh up to around 300lbs each, and the maximum floor tile capacity for each individual cabinet or rack must be kept in mind.  For the deep node styles (H7000, H5600, A3000 and A2000), the considerable node weight may prevent racks from being fully populated with PowerScale equipment. If the cluster uses a variety of node types, installing the larger, heavier nodes at the bottom of each rack and the lighter chassis at the top can help distribute weight evenly across the cluster racks’ floor tiles.

Note that there are no lift handles on the PowerScale 4RU chassis. However, the drive sleds can be removed to provide handling points if no lift is available. With all the drive sleds removed, but leaving the rear compute modules inserted, the chassis weight drops to a more manageable 115lbs or so. It is strongly recommended to use a lift for installation of 4RU chassis.

Cluster backend switches ship with the appropriate rails (or tray) for proper installation of the switch in the rack. These rail kits are adjustable to fit NEMA front rail to rear rail spacing ranging from 22 in to 34 in.

Note that some manufacturers Ethernet switch rails are designed to overhang the rear NEMA rails, helping to align the switch with the PowerScale chassis at the rear of the rack. These require a minimum clearance of 36 in from the front NEMA rail to the rear of the rack, in order to ensure that the rack door can be closed.

Consider the following large cluster topology, for example:

This contiguous rack architecture is designed to scale up to the current maximum PowerScale cluster size of 252 nodes, in 63 4RU chassis, across nine racks as the environment grows – while still keeping cable management relatively simple. Note that this configuration assumes 1RU per node. If using F900 nodes, which are 2RU in size, additional rack capacity should be budgeted for.

Successful large cluster infrastructures depend heavily on the proficiency of the installer and their optimizations for maintenance and future expansion. Some good data center design practices include:

• Pre-allocating and reserving adjacent racks in the same isle to fully accommodate the anticipated future cluster expansion
• Reserving an empty ‘mailbox’ slot in the top half of each rack for any pass-through cable management needs.
• Dedicating one of the racks in the group for the back-end and front-end distribution/spine switches – in this case rack R3.

For Hadoop workloads, PowerScale clusters are compatible with the rack awareness feature of HDFS to provide balancing in the placement of data. Rack locality keeps the data flow internal to the rack.

Excess cabling can be neatly stored in 12” service coils on a cable tray above the rack, if available, or at the side of the rack as illustrated below.

The use of intelligent power distribution units (PDUs) within each rack can facilitate the remote power cycling of nodes, if desired.

For deep nodes such as the H7000 and A3000 hardware, where chassis depth can be a limiting factor, horizontally mounted PDUs within the rack can be used in place of vertical PDUs, if necessary. If front-mounted, partial depth Ethernet switches are deployed, horizontal PDUs can be installed in the rear of the rack directly behind the switches to maximize available rack capacity.

With copper cables (SFP+, QSFP, CX4, etc) the maximum cable length is typically limited to 10 meters or less. After factoring in for dressing the cables to maintain some level of organization and proximity within the racks and cable trays, all the racks with PowerScale nodes need to be in close physical proximity to each other –either in the same rack row or close by in an adjacent row – or adopt a leaf-spine topology, with leaf switches in each rack.

If greater physical distance between nodes is required, support for multimode fiber (QSFP+, MPO, LC, etc) extends the cable length limitation to 150 meters. This allows nodes to be housed on separate floors or on the far side of a floor in a datacenter if necessary. While solving the floor space problem, this does have the potential to introduce new administrative and management challenges.

The various cable types, form factors, and supported lengths available for PowerScale nodes:

Cable Form Factor	Medium	Speed (Gb/s)	Max Length
QSFP28	Optical	100Gb	30M
MPO	Optical	100/40Gb	150M
QSFP28	Copper	100Gb	5M
QSFP+	Optical	40Gb	10M
LC	Optical	25/10Gb	150M
QSFP+	Copper	40Gb	5M
SFP28	Copper	25Gb	5M
SFP+	Copper	10Gb	7M
CX4	Copper	IB QDR/DDR	10M

The connector types for the cables above can be identified as follows:

As for the nodes themselves, the following rear views indicate the locations of the various network interfaces:

Note that Int-a and int-b indicate the primary and secondary back-end networks, whereas Ext-1 and Ext-2 are the front-end client networks interfaces.

Be aware that damage to the InfiniBand or Ethernet cables (copper or optical fibre) can negatively affect cluster performance. Never bend cables beyond the recommended bend radius, which is typically 10–12 times the diameter of the cable. For example, if a cable is 1.6 inches, round up to 2 inches and multiply by 10 for an acceptable bend radius.

Cables differ, so follow the explicit recommendations of the cable manufacturer.

The most important design attribute for bend radius consideration is the minimum mated cable clearance (Mmcc). Mmcc is the distance from the bulkhead of the chassis through the mated connectors/strain relief including the depth of the associated 90 degree bend. Multimode fiber has many modes of light (fiber optic) traveling through the core. As each of these modes moves closer to the edge of the core, light and the signal are more likely to be reduced, especially if the cable is bent. In a traditional multimode cable, as the bend radius is decreased, the amount of light that leaks out of the core increases, and the signal decreases. Best practices for data cabling include:

• Keep cables away from sharp edges or metal corners.
• Avoid bundling network cables with power cables. If network and power cables are not bundled separately, electromagnetic interference (EMI) can affect the data stream.
• When bundling cables, do not pinch or constrict the cables.
• Avoid using zip ties to bundle cables, instead use velcro hook-and-loop ties that do not have hard edges, and can be removed without cutting. Fastening cables with velcro ties also reduces the impact of gravity on the bend radius.

Note that the effects of gravity can also decrease the bend radius and result in degradation of signal power and quality.

Cables, particularly when bundled, can also obstruct the movement of conditioned air around the cluster, and cables should be secured away from fans, etc. Flooring seals and grommets can be useful to keep conditioned air from escaping through cable holes. Also ensure that smaller Ethernet switches are drawing cool air from the front of the rack, not from inside the cabinet. This can be achieved either with switch placement or by using rack shelving.

In this article, we turn our attention to some of the environmental and logistical aspects of cluster design, installation and management.

In addition to available rack space and physical proximity of nodes, provision needs to be made for adequate power and cooling as the cluster expands. New generations of drives and nodes typically deliver increased storage density, which often magnifies the power draw and cooling requirements per rack unit.

The recommendation is for a large cluster’s power supply to be fully redundant and backed up with a battery UPS and/or power generator. In the worst instance, if a cluster does loose power, the nodes are protected internally by filesystem journals which preserve any in-flight uncommitted writes. However, the time to restore power and bring up a large cluster from an unclean shutdown can be considerable.

Like most data center equipment, the cooling fans in PowerScale nodes and switches pull air from the front to back of the chassis. To complement this, data centers often employ a hot isle/cold isle rack configuration, where cool, low humidity air is supplied in the aisle at the front of each rack or cabinet either at the floor or ceiling level, and warm exhaust air is returned at ceiling level in the aisle to the rear of each rack.

Given the significant power draw, heat density, and weight of cluster hardware, some datacenters are limited in the number of nodes each rack can support. For partially filled racks, the use of blank panels to cover the front and rear of any unfilled rack units can help to efficiently direct airflow through the equipment.

The table below shows the various front and back-end network speeds and connector form factors across the PowerScale storage node portfolio.

Speed (Gb/s)	Form Factor	Front-end/ Back-end	Supported Nodes
100/40	QSFP28	Back-end	F900, F600, H700, H7000, A300, A3000, P100, B100
40 QDR	QSFP+	Back-end	F800, F810, H600, H5600, H500, H400, A200, A2000
25/10	SFP28	Back-end	F900, F600, F200, H700, H7000, A300, A3000, P100, B100
10 QDR	QSFP+	Back-end	H400, A200, A2000
100/40	QSFP28	Front-end	F900, F600, H700, H7000, A300, A3000, P100, B100
40 QDR	QSFP+	Front-end	F800, F810, H600, H5600, H500, H400, A200, A2000
25/10	SFP28	Front-end	F900, F600, F200, H700, H7000, A300, A3000, P100, B100
25/10	SFP+	Front-end	F800, F810, H600, H5600, H500, H400, A200, A2000
10 QDR	SFP+	Front-end	F800, F810, H600, H5600, H500, H400, A200, A2000

With large clusters, especially when the nodes may not be racked in a contiguous manner, having all the nodes and switches connected to serial console concentrators and remote power controllers is highly advised. However, to perform any physical administration or break/fix activity on nodes you must know where the equipment is located and have administrative resources available to access and service all locations.

As such, the following best practices are recommended:

• Develop and update thorough physical architectural documentation.
• Implement an intuitive cable coloring standard.
• Be fastidious and consistent about cable labeling.
• Use the appropriate length of cable for the run and create a neat 12” loop of any excess cable, secured with Velcro.
• Observe appropriate cable bend ratios, particularly with fiber cables.
• Dress cables and maintain a disciplined cable management ethos.
• Keep a detailed cluster hardware maintenance log.
• Where appropriate, maintain a ‘mailbox’ space for cable management.

Disciplined cable management and labeling for ease of identification is particularly important in larger PowerScale clusters, where density of cabling is high. Each chassis can require up to twenty eight cables, as shown in the table below:

Cabling Component	Medium	Cable Quantity per Chassis
Back end network	Ethernet or Infiniband	8
Front end network	Ethernet	8
Management Interface	1Gb Ethernet	4
Serial Console	DB9 RS 232	4
Power cord	110V or 220V AC power	4
Total		28

The recommendation for cabling a PowerScale chassis is as follows:

• Split cabling in the middle of the chassis, between nodes 2 and 3.
• Route Ethernet and Infiniband cables towards lower side of the chassis.
• Connect power cords for nodes 1 and 3 to PDU A and power cords for nodes 2 and 4 to PDU B.
• Bundle network cables with the AC power cords for ease of management.
• Leave enough cable slack for servicing each individual node’s FRUs.

Similarly, the stand-alone F-series all flash nodes, in particular the 1RU F600 and F200 nodes, also have a similar density of cabling per rack unit:

Cabling Component	Medium	Cable Quantity per F-series node
Back end network	10 or 40 Gb Ethernet or QDR Infiniband	2
Front end network	10 or 40Gb Ethernet	2
Management Interface	1Gb Ethernet	1
Serial Console	DB9 RS 232	1
Power cord	110V or 220V AC power	2
Total		8

Consistent and meticulous cable labeling and management is particularly important in large clusters. PowerScale chassis that employ both front and back end Ethernet networks can include up to twenty Ethernet connections per 4RU chassis.

In each node’s compute module, there are two PCI slots for the Ethernet cards (NICs). Viewed from the rear of the chassis, in each node the right hand slot (HBA Slot 0) houses the NIC for the front end network, and the left hand slot (HBA Slot 1) the NIC for the front end network. In addition to this, there is a separate built-in 1Gb Ethernet port on each node for cluster management traffic.

While there is no requirement that node 1 aligns with port 1 on each of the backend switches, it can certainly make cluster and switch management and troubleshooting considerably simpler. Even if exact port alignment is not possible, with large clusters, ensure that the cables are clearly labeled and connected to similar port regions on the backend switches.

PowerScale nodes and the drives they contain have identifying LED lights to indicate when a component has failed and to allow proactive identification of resources. The ‘isi led’ CLI command can be used to proactive illuminate specific node and drive indicator lights to aid in identification.

Drive repair times depend on a variety of factors:

• OneFS release (determines Job Engine version and how efficiently it operates)
• System hardware (determines drive types, amount of CPU and RAM, etc)
• Filesystem: Amount of data, data composition (lots of small vs large files), protection, tunables, etc.
• Load on the cluster during the drive failure

A useful method to estimate future FlexProtect runtime is to use old repair runtimes as a guide, if available.

The drives in the PowerScale chassis-based platforms have a bay-grid nomenclature, where A-E indicates each of the sleds and 0-6 would point to the drive position in the sled. The drive closest to the front is 0, whereas the drive closest to the back is 2/3/5, depending on the drive sled type.

When it comes to updating and refreshing hardware in a large cluster, swapping nodes can be a lengthy process of somewhat unpredictable duration. Data has to be evacuated from each old node during the Smartfail process prior to its removal, and restriped and balanced across the new hardware’s drives. During this time there will also be potentially impactful group changes as new nodes are added and the old ones removed.

However, if replacing an entire node-pool as part of a tech refresh, a SmartPools filepool policy can be crafted to migrate the data to another nodepool across the back-end network. When complete, the nodes can then be Smartfailed out, which should progress swiftly since they are now empty.

If multiple nodes are Smartfailed simultaneously, at the final stage of the process the node remove is serialized with around 60 seconds pause between each. The Smartfail job places the selected nodes in read-only mode while it copies the protection stripes to the cluster’s free space. Using SmartPools to evacuate data from a node or set of nodes in preparation to remove them is generally a good idea, and is usually a relatively fast process.

Another efficient approach can often be the swapping out of drives into new chassis. In addition to being considerable faster, the drive swapping process focuses the disruption on a single whole cluster down event. Estimating the time to complete a drive swap, or ‘disk tango’ process is simpler and more accurate and can typically be completed in a single maintenance window.

With PowerScale chassis-based platforms, such as the H700 and A300, the available hardware ‘tango’ options are expanded and simplified. Given the modular design of these platforms, the compute and chassis tango strategies typically replace the disk tango:

Replacement Strategy	Component	PowerScale F-series	Chasis-based nodes	Description
Disk tango	Drives / drive sleds			Swapping out data drives or drive sleds
Compute tango	Chassis Compute modules			Rather than swapping out the twenty drive sleds in a chassis, it’s usually cleaner to exchange the four compute modules
Chassis tango	4RU Chassis			Typically only required if there’s an issue with the chassis mid-plane.

Note that any of the above ‘tango’ procedures should only be executed under the recommendation and supervision of Dell support.