We try to be optimal while scanning through large rows with many version that are not wanted, but nothing is as optimal as only storing minimal data and retrieving that. We'd love to hear about any benchmarks or speed runs you have.
Regards, -ryan On Mon, Dec 6, 2010 at 11:15 PM, Hari Sreekumar <[email protected]> wrote: > Hi Stack, > > Thanks a lot for pointing out, I had overlooked it I went through the > API. My data's no. of versions is highly variable. Basically, I am trying to > use one row for a user's browser session and creating a new version for each > user action (click etc.), storing data like where the user clicked. This > means a row can have just one version, or it can even have thousands of > versions. I am thinking of getting all versions using the time range > option. But is that efficient? And what are the factors to be kept in mind > in this case? Will it be efficient to have some rows with 1 version and some > with 100 versions? > > Thanks again, > Hari > > On Tue, Dec 7, 2010 at 12:22 PM, Stack <[email protected]> wrote: > >> You have to 'Get' all or N versions. See how >> >> http://hbase.apache.org/docs/r0.89.20100924/apidocs/org/apache/hadoop/hbase/client/Scan.html >> and >> http://hbase.apache.org/docs/r0.89.20100924/apidocs/org/apache/hadoop/hbase/client/Get.html >> allow you stipulate how many versions to return (You can also specify >> a TimeRange). >> >> St.Ack >> >> >> On Mon, Dec 6, 2010 at 10:16 PM, Hari Sreekumar >> <[email protected]> wrote: >> > Hi, >> > >> > How is versioning used? I cannot know all the timestamps corresponding >> to >> > each column in my design. But I need to use the older versions of >> columns. >> > How can this be done? Is there a way to output all previous versions of a >> > column/column-family? >> > >> > Hari >> > >> >
