Home > Storage > PowerScale (Isilon) > Product Documentation > Management and Migration > Storage Tiering with Dell PowerScale SmartPools > Caching
SmartCache is a globally coherent read and write caching infrastructure that provides low latency access to content. Like other resources in the cluster, as more nodes are added, the total cluster cache grows in size, enabling OneFS to deliver predictable, scalable performance within a single file system.
A OneFS cluster provides a high cache to disk ratio (multiple GB per node), which is dynamically allocated for read operations as required. This cache is unified and coherent across all nodes in the cluster, allowing a user on one node to benefit from I/O already transacted on another node. OneFS stores only distinct data on each node. The node’s RAM is used as a level 2 (L2) cache of such data. These distinct, cached blocks can be quickly accessed across the backplane, and, as the cluster grows, the cache benefit increases. For this reason, the amount of I/O to disk on a OneFS cluster is most often quite lower than on traditional platforms, allowing reduced latencies and a better user experience. For sequentially accessed data, OneFS SmartRead aggressively prefetches data, greatly improving read performance across all protocols.
An optional third tier of read cache, called SmartFlash or level 3 (L3) cache, is also configurable on nodes that contain solid state drives (SSDs). SmartFlash is an eviction cache that L2 cache blocks populate as they are aged out from memory. Using SSDs for caching rather than as traditional file system storage devices offers several benefits. For example, when reserved for caching, the entire SSD is used, and writes occur in a linear and predictable way. Using SSDs for caching provides far better utilization and also results in considerably reduced wear and increased durability over regular file system usage, particularly with random write workloads. Using SSDs for cache also makes sizing SSD capacity a more straightforward and less error-prone prospect compared to using SSDs as a storage tier.
OneFS write caching uses write buffering to aggregate, or coalesce, multiple write operations to the NVRAM file systems journals. The operations can be written to disk safely and more efficiently. This form of buffering reduces the disk write penalty that could result in multiple reads and writes for each write operation.
Note: For more information, see the OneFS SmartFlash white paper.