Disk I/O is a major component of any database system, and Elasticsearch is no exception. Elasticsearch, along with most modern databases, makes heavy use of filesystem cache (RAM) to optimize performance. However, when persistence is needed — and it’s needed often — the disk is used. Heavy disk use under a high load can cause a bottleneck with I/O wait, which is a big factor in performance loss.
That being said, it can’t hurt to have blazing fast hard drives as a foundation. With Elasticsearch, faster disk writes = faster indexing, faster disk reads = faster searching.
Since launching our dedicated service, Qbox had only offered SSD as external storage on Rackspace. As of yesterday, however, we now provide 12 new node sizes with SSD on both Amazon EC2 and Rackspace clouds. We’ve also kept the non-SSD t1.micro and t1.small on EC2 as well as the 512MB standard instance on Rackspace for testing purposes and smaller deployments.
Qbox customers with existing non-SSD clusters can continue to add/remove nodes as normal.
But if you ever need a performance boost, check out a cluster on SSD. We’re more than happy to migrate your data if you choose to go with Qbox SSD.