I think any ticket that results in less work being done is important, even if it's small, because the accumulation of these savings can have significant impact.
I can't comment on how a lot of these may impact users like myself, but I'd like to echo that this is indeed a severe pain point that I recently ran into while trying to improve write rates and index updates was the single biggest limiting factor. For example, I have a table with two indices on it and the columns in the index are only comprised of columns that make up the PK. Writes are done using K/V APIs and on average only about 15% of the writes would involve creating a new key, the rest are updating the value only. At a write rate of about 60k/sec (to a node consisting of 16 cores) the cpu utilization is at about 60%. However, if I remove the indices the cpu utilization drops to about 30%. That's really significant and unexpected. AFAICT this scenario will largely benefit from https://issues.apache.org/jira/browse/IGNITE-7015. I think for any workload that's heavy on writes and requires low latency reads against large data sets any performance improvement you can make to index updates/traversal/etc is a huge huge win. -Nick -- Sent from: http://apache-ignite-developers.2346864.n4.nabble.com/
