Benchmarking a database designed for multiple nodes on a single node usually doesn't give you any measurements that are useful. We recommend 5 nodes minimum for clusters, so if you can do that, it is best to benchmark on a realistic cluster. If you are determined to benchmark on a single node, you should configure your N value to be 1 so you aren't doing 3x the work. Also, you should probably set your ring size to 8 so you don't run out of memory (which I suspect is your problem).
I replied similarly here http://riak-users.197444.n3.nabble.com/How-to-make-Riak-work-faster-writing-tp4025882p4025909.htmland you can read that to see how to adjust the n_val on a bucket. Good luck. -Jared On Wed, Sep 25, 2013 at 9:50 AM, Wagner Camarao <wag...@crunchbase.com>wrote: > Hi, > > I'm benchmarking 2i at scale of billion records, running one physical node > locally with mostly default configs - except for LevelDB instead of > Bitcask. Up to this point (14MM records in the bucket that's being indexed) > it's still performing lookups well for my use case (read ~ 7ms using > riak-ruby-client over http). > > However, along this process I've noticed riak to go down twice. First time > (8MM records) I could just start it again and continue my benchmarking from > the point it were left, but now at the second time (14MM records) when I > started riak again, it took about 3 minutes to respond to my first request. > > What was happening during these long startup minutes, after my second > crash? > > Up to which scale have you guys been successfully using secondary indexes? > > Any other ideas given my use case / benchmarking scenario? > > Thanks, > Wagner > > _______________________________________________ > riak-users mailing list > riak-users@lists.basho.com > http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com > >
_______________________________________________ riak-users mailing list riak-users@lists.basho.com http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
