CloudPhysics: Storage Cache Analysis Reinvented with SHARDS
We’re excited to share the work that was recently released at the presitgious USENIX FAST ’15 conference. Carl Waldspurger presented “Efficient MRC Construction with SHARDS,” a paper that he coauthored with other CloudPhysics technologists Nohhyun Park, Alexander Garthwaite, and me. Click below view the talk (video length is 25 minutes).
For those that prefer a quick synopsis of this exploration of our storage cache analysis:
Caches are pervasive in modern storage systems.
Designed to accelerate data access by exploiting locality, a cache provides an essential service in modern computing. Operating systems and databases maintain in-memory buffer caches containing hot blocks. Here, a hot block is one that is considered likely to be reused. Storage cache using flash memory are popular as a cost-effective way to reduce application latency and offload work from rotating disks. This has recently been a trend both in networked as well as server-side storage. Today, virtually all storage devices — ranging from individual disk drives to large storage arrays — include significant caches composed of RAM or flash memory.
Read the entire article here, Storage Cache Analysis Reinvented with SHARDS