It's touchy... what if the data set doesn't fit in the in-memory's part of the block cache (which is 25%)? Maybe the user only wants to keep "in-memory" those edits that are being used? What about the IO hit of assigning those regions at startup that would now need to read X GBs all at once?
FWIW I've never been a fan of that setting. J-D On Tue, Feb 26, 2013 at 5:04 PM, Sergey Shelukhin <[email protected]> wrote: > should we make this built-in? Sounds like default user intent for in-memory. > > On Tue, Feb 26, 2013 at 2:13 PM, Stack <[email protected]> wrote: > >> On Fri, Feb 22, 2013 at 7:39 AM, Eric Czech <[email protected]> wrote: >> >> > Hi everyone, >> > >> > Are blocks for in-memory column families automatically loaded in to the >> > block cache on restart? >> >> >> No >> >> >> > If not, would anyone recommend running a scan with >> > .setCacheBlocks(true) after a restart for in-memory column families? >> > >> >> Yes. >> >> It should be easy verifying whether the above warmup had an effect. >> >> Good luck, >> St.Ack >>
