You are viewing a plain text version of this content. The canonical link for it is here.
Posted to notifications@skywalking.apache.org by GitBox <gi...@apache.org> on 2019/03/29 03:45:15 UTC

[GitHub] [incubator-skywalking] mbao commented on issue #2271: Improve OAP cluster performance

mbao commented on issue #2271: Improve OAP cluster performance 
URL: https://github.com/apache/incubator-skywalking/pull/2271#issuecomment-477856006
 
 
   > In these days tests, by my back to back works with @hanahmily , we made big progresses in
   > 
   > 1. Reduce CPU cost in low payload
   > 2. Improve single node in 8U(GCP) performance to 80k telemetry per second.
   > 
   > But we still need to work on improving performance in cluster mode and when scale out.
   > 
   > Right now, 3 nodes cluster has only the same performance as a single node, also scaling out from 1 to 3 doesn't make performance better.
   > 
   > In this pull request, I do following changes
   > 
   > 1. Remove the codes about sending to L2 in every 1000 indicator rows.
   > 2. Make L2 send based on the timeline, send every second by default.
   > 3. Settings related to DataCarrier channel and buffer sizes could be set by System env.
   > 
   > Let's see the test results.
   
   Do you mean single node can handle as much segment as a cluster of three nodes?
   I have a cluster of two nodes, when there is only one node, CPM value is the same as grep count from access count, but when I add one node, CPM is smaller than the count.

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services