hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Nicolas Liochon (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-9775) Client write path perf issues
Date Thu, 07 Nov 2013 09:03:17 GMT

    [ https://issues.apache.org/jira/browse/HBASE-9775?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13815782#comment-13815782

Nicolas Liochon commented on HBASE-9775:

bq.  To be checked.
I confirm. That's by design (in YCSB, and in a way, in HBase as well, 0.94 included...). It
does not really match a real use case, an application server would have a single connection.

32 clients on a single machine takes some CPU (whatever the number of thread), I guess with
all the changes we're likely better, but we could be CPU bound.

A test with a single YCSB client on each machine would be interesting imho.

bq. Ideally we want something like what you had before – 5 or 1/2 the CPUs on the local
server as guesstimate of how many CPUs the server has, which ever is greater
IIRC, I forgot to change hbase-site.xml, so the default has not changed at the end :-).
But it's a per AsyncProcess limit. With the 32 clients, a server can have 64 writes  queries
in parallel. That's already a lot. (with 0.94, it would have been 32 max, and much likely
half of less of it, as we wait for all the queries to be finished before sending a new set
of queries). 

> Client write path perf issues
> -----------------------------
>                 Key: HBASE-9775
>                 URL: https://issues.apache.org/jira/browse/HBASE-9775
>             Project: HBase
>          Issue Type: Bug
>          Components: Client
>    Affects Versions: 0.96.0
>            Reporter: Elliott Clark
>            Priority: Critical
>         Attachments: 9775.rig.txt, 9775.rig.v2.patch, 9775.rig.v3.patch, Charts Search
  Cloudera Manager - ITBLL.png, Charts Search   Cloudera Manager.png, hbase-9775.patch, job_run.log,
short_ycsb.png, ycsb.png, ycsb_insert_94_vs_96.png
> Testing on larger clusters has not had the desired throughput increases.

This message was sent by Atlassian JIRA

View raw message