Map/reduce aside, in the general case, I do time series in Riak with deterministic materialized keys at specific time granularities. Ie.
/devices/deviceID_YYYYMMDDHHMM[SS] So my device or app stack will drop data into a one second resolution key (if second resolution is needed) into Riak memory backend in a materialized key (the combination of ID and timestamp). Then retrieve the data deterministically (so you skip the search) at a one minute resolution by issuing a 60 key multi get for each second in that minute. Have a trailing process that sweeps memory and does larger granularity rollups/analytics/math/whatever and write that to memory or disk. All depends on frequency. Best, Alexander @siculars http://siculars.posthaven.com Sent from my iRotaryPhone > On Feb 19, 2015, at 19:08, Christopher Meiklejohn <cmeiklej...@basho.com> > wrote: > > >> On Feb 19, 2015, at 8:01 PM, Fred Grim <fg...@vmware.com> wrote: >> >> Given a specific data blob I want to move a time series into a search >> bucket. So first I have to build out the time series and then move it >> over. Maybe I should use the rabbitmq post commit hook to send the data >> somewhere else for the query to be run or something like that? > > Given your scenario, it seems that a portion of these writes would have > MapReduce jobs that resulted in nothing happening — I assume you only > bucket the series every so many writes or time period, correct? > > I’d highly recommend doing this externally, or identifying a method for > pre-bucket’ing the data given the rate of ingestion. > > - Chris > > Christopher Meiklejohn > Senior Software Engineer > Basho Technologies, Inc. > cmeiklej...@basho.com > _______________________________________________ > 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
