[docdb] Consider storing data on yb-master in an uncompressed format #5170
Labels
area/docdb
YugabyteDB core features
kind/enhancement
This is an enhancement of an existing feature
priority/medium
Medium priority issue
Projects
Jira Link: DB-1900
While running a large TPCC workload (sharded 10 ways) on a 30 node cluster, the CPU usage of the YB-Master leader spiked up for a few secs/mins every time a new TPCC benchmark shard was started. Because of the CPU load. a
perf record
was failing. But upon investigating further usingperf top -p <YB Master leader PID>
, a disproportionate amount of CPU (almost 17%) seemed to be going towardssnappy::RawUncompress
. This is shown in the screen capture shown below.A simple optimization could be to not compress the data in YB Master in order to make it more CPU efficient
.
The text was updated successfully, but these errors were encountered: