Discover Technology: EMC - iSilon

Useful link:

EMC iSilon SmartPool Basic Information

Isilon OneFS 7.1: SmartPools Lab Setup

SmartPools is a license feature of OneFS that provides the ability to manage the performance and cost of data in a cluster as it relates to business objectives such as value and SLAs. All SmartPools configuration is performed on the OneFS cluster as a single point of management which simplifies administration.

SmartPools provides the ability to build a cluster using different node types in different pools and tiers to meet those business objectives.

SmartPools is policy driven rather than heuristic which allows for deterministic alignment of data to a given tier. These policies, called File Pool policies, consider whole files for migration between tiers based on filesystem metadata. Only the underlying blocks pertaining to a file are moved, the file stays in the same place from a filesystem perspective. Policies are often implemented to provide a specific type of performance and/or cost objective to a given file or dataset.

SmartPools runs as a background job to migrate data from one tier to another as needed. SmartPools file pool policies can also be implemented to ingest all data written to a defined path to a specific tier of storage.

SmartPools helps to align business and technical requirements for cost and performance to file data using policies. Managing the growth of tiers of storage in a dynamic way helps to contain costs as the business expands its capacity. Migrating data with SmartPools as it ages allows for even better alignment of the data with its value for the business.
Module

Many customers will implement different node types as clusters increase in size helping to differentiate the performance and cost metrics associated with capacity and SmartPools is required to make that work intelligently.

Large scale clusters will often times include different node types to minimize costs for infrequently accessed content. The ability to move data between these tiers is a key benefit and feature of SmartPools.

SLAs between business units and applications in terms of cost per gigabyte and cost per performance may dictate that different storage nodes are used to service those needs. Many customers consider these SLAs in terms of bronze, silver, and gold levels.

Policies are in place to move data that has been inactive for 30 days to the NL-series tier to minimize the cost associated with storing that inactive data.

Active files greater than 2GB have been targeted with a policy to be moved to the X-series tier since those files are likely good candidates for SATA performance with large sequential I/O.

The default policy writes all new data to the S-series tier to provide the highest possible performance for ingest and highly active data.

Each night SmartPools considers which files need to move based on the policies in place and will perform those migrations in the background

SmartPools groups components within a cluster into different logical units. The smallest unit an administrator can manage is the Node Pool.

Node pools group similar nodes within a cluster automatically according to defined compatibility rules

There is a minimum requirement of three nodes for a supported node pool, and a pool can be as large as the maximum supported cluster size, currently 144 nodes.

Specific Node pools can be used as a target for SmartPools file pool policies

•SmartPools will also subdivide disks in node pools to provide higher fault tolerance while minimizing the protection overhead associated with that fault tolerance.
•SmartPools provisions these disk pools automatically according to defined rules and is not configurable by an administrator.

•SmartPools allows an administrator to define tiers as a SmartPools file pool policy storage target. This provides greater flexibility in configuring SmartPools in an environment.
•Many customers with different generations or capacities in a given tier will consider using tiers as a way to group nodes with sufficient performance for an application into a single target rather than having to manage more complex policies to distribute that data across many node pools.
•A node pool can only exist in a single tier at any time

•File pool policies can be named to help with the administration of policies and a description can be added.

•Filters are applied that will match against file attributes to select files. Those attributes are;

•File name

•Path

•File type

•Timestamps

•Custom user metadata

•Policies can be marked to stop processing additional rules to ensure their changes are the last affected on a file.

•SmartPools File Pool Policies are used to define how data is ingested into the cluster and how that data migrates between tiers.

•Policies are built using a collection of file characteristic filters that are used to select files for migration or transformation. Policies also define how that data is changed be it directed to a different tier or node pool, its protection level modified, or I/O optimization settings changed.

•The migration of data between tiers is completely transparent to users—they will still access the file in its original path and will not be directed to some other subsystem for its access.