+1 for redis for this use case. On Aug 16, 2013, at 10:54 AM, Robert Coli <rc...@eventbrite.com> wrote:
> On Fri, Aug 16, 2013 at 10:43 AM, Todd Nine <tn...@apigee.com> wrote: > We're using expiring columns as a mean for locking. > > Perhaps a log structured data store with immutable data files is not ideal > for your use case? > > If I were you, I'd put this use case in Redis and be done with it instead of > trying to get Cassandra to do something that is the opposite of what it is > optimized for. > > However, we're still seeing very wide rows in our sstables, which (I'm > assuming) are due to tombstones since I only get < 10 columns back on > a full range scan. > > Don't assume, use tracing to show you how many tombstones are being read? > > We're trying to eliminate the need to go to disk > completely in the Locks\HLocks CF. Is there anything further we can > do? > > populate_io_cache_on_flush option may help you. > > =Rob
