Strange merge time variance

David Smith dizzyd at basho.com
Mon Dec 19 16:13:41 EST 2011


On Thu, Dec 15, 2011 at 10:51 AM, Dmitry Groshev <lambdadmitry at gmail.com> wrote:
> I'm using standalone bitcask as a backend for our in-memory database
> (not opensourced yet). I write ~1.5GB of data to it each 5 minutes
> overwriting almost all keys and after that I run a merge as follows:

If the database is in memory, why write it to bitcask? Seems like a
simple replay log is what you want in this situation.


> depicts a durations of merges; as one can see, there at least 2 "saw"
> forms with different time scales overlapped. Is there any reason why
> it is so? I have a gut feeling that such a variance isn't so good :)

Could this be an artifact of the large # of writes + deletes on the
file system? Otherwise, I'm not quite sure what a good pattern would
look like...

D.

-- 
Dave Smith
Director, Engineering
Basho Technologies, Inc.
dizzyd at basho.com



More information about the riak-users mailing list