[newbie] riak performance - basics

Dan Reverri dan at basho.com
Wed Apr 13 11:24:05 EDT 2011


Hi MH,

Are there any errors in the Riak logs? I am wondering if some of those
requests are failing, possibly due to an open file handle limit:
https://wiki.basho.com/Open-Files-Limit.html

Thanks,
Dan

Daniel Reverri
Developer Advocate
Basho Technologies, Inc.
dan at basho.com


On Wed, Apr 13, 2011 at 7:38 AM, Kelly McLaughlin <kelly at basho.com> wrote:

> MH,
>
> Riak may not seem as fast in a single node configuration compared to things
> like mongodb or others, but keep in mind the strengths of Riak are its
> performance at scale and the ease with which it does scale. Nobody runs a
> single node configuration in an actual production environment so keep that
> in mind as you draw conclusions from your tests. I would encourage you to do
> some testing on a cluster of nodes. Riak is written in Erlang and the
> benefits of Erlang really become apparent when there is a lot of concurrent
> load. The advantages of Riak specifically become apparent when you have a
> few nodes crash, but don't lose any data and your cluster keeps humming
> along. I would also point out the ease of administration with Riak. Scaling
> either up or down is very simple I think when compared to other systems that
> use shards or other schemes. Hope this helps some. Good luck and have fun!
>
>
> -Kelly
>
>
> On Wed, Apr 13, 2011 at 6:58 AM, M H <jdevmh at gmail.com> wrote:
>
>> Hi,
>>
>> I am new to Riak - just compiled it and trying it out. I have some
>> questions about performance.
>>
>> I am running a single riak node with default configuration (only
>> changed the IPs) on a core i7 machine (Ubuntu). I run a very simple
>> JMeter test (RESTful calls):
>>  1 .PUT a simple json object into riak (~30 bytes; key is static - so
>> DB has only one key)
>>  2. GET the object
>>
>> While the test is running riak seems to be using all the CPU it can
>> get - 'top' reports ~99.5% CPU usage (beam.smp is using ~780% so all 8
>> cores).
>>
>> Here are some results from the JMeter test:
>> * 5 threads:
>>  1.PUT  avg: 458 ms  Throughput 6.1/sec
>>  2.GET  avg: 355 ms  Throughput 6.1/sec
>>
>> * 50 threads:
>>  1.PUT  avg: 5325 ms  Throughput 6.1/sec
>>  2.GET  avg: 2838 ms  Throughput 6.1/sec
>>
>> The numbers are pretty disappointing. Is there something wrong with
>> the default configuration? Or I am misusing riak in some way? There is
>> only 1 key with simple value in the DB....
>>
>>
>> Here's some numbers from admin stats:
>>
>> vnode gets : 2118
>> vnode_puts : 1067
>> read_repairs : 1
>> vnode_gets_total : 224061
>> vnode_puts_total : 89045
>> node_gets : 680
>> node_gets_total : 74676
>> node_get_fsm_time_mean : 3047570
>> node_get_fsm_time_median : 2949500
>> node_get_fsm_time_95 : 4888000
>> node_get_fsm_time_99 : 5964050
>> node_get_fsm_time_100 : 6348870
>> node_puts : 339
>> node_puts_total : 29543
>> node_put_fsm_time_mean : 2433651
>> node_put_fsm_time_median : 2472125
>> node_put_fsm_time_95 : 4054512
>> node_put_fsm_time_99 : 4504152
>> node_put_fsm_time_100 : 4908213
>> read_repairs_total : 180
>> cpu_nprocs : 620
>> cpu_avg1 : 2104
>> cpu_avg5 : 2117
>> cpu_avg15 : 2099
>> storage_backend : riak_kv_bitcask_backend
>>
>> Any hints?
>>
>> --
>> MH
>>
>> _______________________________________________
>> riak-users mailing list
>> riak-users at lists.basho.com
>> http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
>>
>
>
> _______________________________________________
> riak-users mailing list
> riak-users at lists.basho.com
> http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.basho.com/pipermail/riak-users_lists.basho.com/attachments/20110413/bfe36a64/attachment.html>


More information about the riak-users mailing list