incubator-blur-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aaron McCurry <amccu...@gmail.com>
Subject Re: Blur Index get much slower when running under multi-node cluster than single cluster
Date Mon, 17 Feb 2014 01:31:20 GMT
First off welcome!  What version are you using?

Sent from my iPad

On Feb 16, 2014, at 7:55 PM, Hualiang Xu <hualiang.xu@gmail.com> wrote:

> Hi All,
> 
> I noticed that blur index speed is much slower when it is under multi-node
> cluster than single cluster. For initial rows added to blur, its speed
> looks good. But as I put more and more rows, the index speed slows down. Do
> anyone see this?

If you are using any version before 0.2.2 (hasn't been released yet) the mutate performance
via thrift is pretty dismal. You can checkout 0.2.2 from the git branch of apache-blur-0.2.
There are a couple of fixes left before the release but I'm hoping to star the vote this week.

Also once you are using 0.2.2, batch mutates can have a huge speed up in performance if that
can fit into your app. 

> 
> If I put all the records in a single row, the index speeds up. However, I
> noticed the query results are row based.

You can run queries to get rows or records. See the query object in thrift to turn off the
row query. 

Hope this helps. 

Aaron

> 
> Thanks for any answer.
> Hualiang

Mime
View raw message