Slowdown in Voldemort 0.80?
Chris Bunch
I've recently upgraded to Voldemort 0.80 from a much older version
(I don't have the version number but it's at least a year old) and
have switched from using the Thrift interface to the native Python
interface using the StoreClient. I've noticed that when I make get,
get_all, put, or delete calls I occasionally get a Protocol Buffer
DecodeError exception (something to the effect of 'truncated
message'). Is this normal?
Furthermore, I'm also seeing quite a bit of a slowdown with this
version compared to the older one we had. While gets take about the
same time, I'm seeing about a 50% slowdown on the put and delete
calls. Do you also see the same?
Thanks,
Alex Feinberg
There's been some changes made to server side routing (the method used
by the Python and other "thin clients") in that time period. The most
recent changes were made to accommodate rebalancing (namely, ability
to keep taking read/write requests as partitions were being migrated
between nodes). You may want to attempt to turn of these properties in
server.properties:
"enable.rebalancing" (set it to false)
"enable.redirecting" (set it to false false)
"enable.metadata.checking" (set it to false)
Let me know if you then see a performance increase after the settings.
If so, that means that there were inefficiencies introduced in how
server-side routing deals with rebalancing related changes (especially
when rebalancing isn't in progress). These should likely be easy to
fix.
In addition, could you paste the full error trace of the exceptions
you see, from both the client and the server. We should definitely not
be seeing these.
Thanks!
- Alex
> --
> You received this message because you are subscribed to the Google Groups "project-voldemort" group.
> To post to this group, send email to project-...@googlegroups.com.
> To unsubscribe from this group, send email to project-voldem...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/project-voldemort?hl=en.
>
>