E-Series SSD Cache refers to SSD drives that are installed in NetApp E-Series storage arrays to accelerate read throughput. The SSDs available in the E-Series are the largest that NetApp offers at 800GB. The minimum amount of cache is a single SSD drive, and the maximum is 5TB of SSD Cache per E-Series array.
E-Series SSD Cache is similar to the Flash Pools used in our FAS and V-Series systems (discussed in this blog) but with some important design distinctions. Like Flash Pool, SSD Cache is used to automatically cache data blocks from HDD to SSD. Both do this in real time, without the need for policy engines or scheduled batch data migration, and both copy, rather than migrate, hot data to cache, with the source data always residing on HDD.
Inside The E-Series SSD Cache
As I mentioned, the E-Series SSD Cache contains some important design distinctions. To understand the differences between E-Series SSD Cache and FAS Flash Pools, it’s helpful to understand the design goals of each storage array.
NetApp FAS storage arrays are designed as general-purpose SAN/NAS enterprise storage systems that support a wide variety of IT workloads. As such, they are designed with agility, efficiency, and scalability in mind.
E-Series, on the other hand, is a purpose-build SAN storage array that is designed for speed, speed and more speed. An E-Series array typically finds itself at home serving high performance data for applications such as real time analytics, OLTP, and streaming video. Therefore, E-Series SSD Cache must be designed for performance optimization in support of these workloads. SSD Cache accomplishes this in several ways:
· E-Series SSD Cache has adjustable parameters that are used to fine-tune the cache. Cache block sizes are adjustable from 2K to 8K. Depending on the pattern of the data being cached, block size can have a great effect on the speed of cache warming – internal tests have shown that a properly tuned E-Series cache will warm 500% faster than an improperly tuned cache. This is particularly important in applications with revolving working sets i.e. data analytics.
· Another aspect of E-Series SSD Cache is that newly written data can either be cached immediately, or sent directly to HDD only. Certain applications rely heavily on read-after-write, while other applications store data for consumption at a later time. Optimizing the use of cache based on the read/write characteristics of the application allows for the maximum availability of cache space when it’s needed.
· Finally, the SSD block cache map is held in DRAM, not HDD, to insure the fastest lookup for determination of cache hits and misses. Workloads we’ve tested with E-Series SSD Cache have shown up to a 700% increase in IOPS over arrays without cache.
Below are some additional resources:
Data Storage Matters,