Hi Paul,

What kind of systems are you running?  With moderate CPU, I can drive tens of 
thousands of operations from the Java client, so I'm not sure it's lack of CPU.

Does the same setup go further with RoadRunner?

Thanks,

Matt

From: asdf9898 <[email protected]<mailto:[email protected]>>
Reply-To: "[email protected]<mailto:[email protected]>" 
<[email protected]<mailto:[email protected]>>
Date: Monday, February 3, 2014 1:04 PM
To: "[email protected]<mailto:[email protected]>" 
<[email protected]<mailto:[email protected]>>
Subject: Re: couchbase scale out performance

Aliaksey,

I think that having a second node is creating CPU demand on localhost.

This is where the YCSB client is running and that is somehow limiting what the 
client can produce.

So I think maybe throughput is determined by what CPU cycles are available to 
the ycsb.

Thanks.

On Sunday, February 2, 2014 9:01:47 PM UTC, Aliaksey Kandratsenka wrote:



On Sat, Feb 1, 2014 at 4:38 PM, asdf9898 <[email protected]<javascript:>> 
wrote:
Okay I will try deleting the bucket and creating and then setting replicas to 
zero

But on the "read" performance : 22,000 records /second: does that seem low?

This number drops to 14,000 with 2 nodes: does that number seem low?

22k is not high for sure. But drop to 14k is indeed something interesting and 
it doesn't match our understanding of how things should be.

Is that with ycsb ported to couchbase or some other benchmark ?


--
You received this message because you are subscribed to the Google Groups 
"Couchbase" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to 
[email protected]<mailto:[email protected]>.
For more options, visit https://groups.google.com/groups/opt_out.


--
Matt Ingenthron
Couchbase, Inc.

-- 
You received this message because you are subscribed to the Google Groups 
"Couchbase" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/groups/opt_out.

Reply via email to