Typically you are not going to be doing a single read larger than
heap, unless you have rogue cell. We have scanner caching to control
number of rows pulled in at once, and then we have batching which will
control the number of columns pulled in at once. If you try to grab
too much data you can expect an OOME on the RS itself.
On Tue, Mar 19, 2013 at 1:34 AM, Henry JunYoung KIM
<[EMAIL PROTECTED]> wrote:
> hi, hbase users.
> maybe this is a dummy question for well-known hbase users.
> if read requests are so many many, data size is bigger than heap size. I mean that data size from requests couldn't be loaded at once on memories.
> What situations could be expected inside of hbase?
> flush memstores?
> firstly hbase accept all requests, and then response datas in order although the responses are so late?
> or deny requests?
Customer Operations Engineer, Cloudera