{"id":4040,"date":"2017-09-22T00:03:55","date_gmt":"2017-09-22T07:03:55","guid":{"rendered":"http:\/\/www.couchbase.com\/blog\/?p=4040"},"modified":"2025-06-13T19:28:37","modified_gmt":"2025-06-14T02:28:37","slug":"plasma-index-performance","status":"publish","type":"post","link":"https:\/\/www.couchbase.com\/blog\/plasma-index-performance\/","title":{"rendered":"Blazing Fast Index Performance with Plasma"},"content":{"rendered":"

In the <\/span>earlier post<\/span><\/a>, we got an opportunity to introduce Plasma as the new performant storage engine that powers Global Secondary Indexes in Couchbase 5.0(beta out!). We saw some of the features of Plasma and how it is designed with the latest hardware trends in mind so that Index Performance is vastly improved.<\/span><\/p>\n

We truly believe that Software has to leverage many of the capabilities of the underlying Hardware that it runs on for it to be successful. \u00a0By re-looking at the problem from scratch, our engineers leveraged lock free data structures, along with numerous other optimisations, and came up with a memory-first design in Plasma. This enables us in offering the best experience to our customers who demand the lowest latencies and the best throughput possible.<\/span><\/p>\n

Comparing ForestDB with Plasma is not essentially an apples-to-apples comparison as we have completely rewritten the storage engine from scratch and re-architected the entire component. \u00a0While ForestDB could see only upto 10-20k mutations per second, Plasma sails through with ease even with close to 100k mutations per second; and this is on markedly much larger datasets as well. Having said that, on some of our in-house performance tests, we observe for secondary index scan:<\/span><\/p>\n