Adam Fuchs 2013-04-15, 19:00
William Slacum 2013-04-15, 19:06
Adam Fuchs 2013-04-15, 19:10
Christopher 2013-04-15, 20:13
Adam Fuchs 2013-04-15, 21:06
-Re: multi-table isolated batch scanner
Keith Turner 2013-04-15, 21:33
On Mon, Apr 15, 2013 at 5:06 PM, Adam Fuchs <[EMAIL PROTECTED]> wrote:
> The desire for isolation stems from the desire to amortize some computation
> over a number of results. Say it takes 5 seconds to compute an intersection
Would increasing the size of the key/value buffer help in your case?
The iterator stack is not torn down until that buffer fills up or the
end of tablet is reached. Are you concerned about the cost of
reconstructing the iterator stack across tablets?
> of a couple of sets within the iterators, and then streaming back the
> results takes a minute or so. If I have to redo the 5 second computation
> many times, as in to support the reconstruction of the iterator tree, then
> that computation may start to dominate my query performance. Primarily,
> this means I need to be able to continue a scan without having to rebuild
> the iterators. Isolation in the scanner has that side effect. Proper
> isolation would be a "nice-to-have", but I can deal with not having it.
> On Mon, Apr 15, 2013 at 4:13 PM, Christopher <[EMAIL PROTECTED]> wrote:
>> It seems like you're talking about two features at once:
>> 1) Multi-table batch scanner.
>> 2) Scan Isolation on batch scanners like we have on regular scanners.
>> Is that correct?
>> I can see the utility of a multi-table batch scanner, but I haven't
>> seen a compelling need for implementing isolation on the
>> batch-scanners. Do you have a use case in mind for that?
>> Also, it seems that your use case for isolation is not so much the
>> isolated reads, but the statefulness of the iterator stack on the
>> server side. Is this correct? If so, I'm even more curious about your
>> use case for this, since that statefulness is only guaranteed per-row.
>> Christopher L Tubbs II
>> On Mon, Apr 15, 2013 at 3:10 PM, Adam Fuchs <[EMAIL PROTECTED]> wrote:
>> > Thanks Bill,
>> > I care about latency and throughput. First available result ordering is
>> > fine, though.
>> > Does Guava just chain through a collection of iterators, completing one
>> > then moving to the next?
>> > Adam
>> > On Mon, Apr 15, 2013 at 3:06 PM, William Slacum <
>> > [EMAIL PROTECTED]> wrote:
>> >> How are you expecting to get results back? Guava's Iterables could
>> concat a
>> >> bunch of a Scanners together, if you didn't care about the throughput
>> >> aspect of it and simply wanted results from multiple tables.
>> >> On Mon, Apr 15, 2013 at 3:00 PM, Adam Fuchs <[EMAIL PROTECTED]> wrote:
>> >> > Is anyone else pining for a multi-table isolated batch scanner, or is
>> >> > just me? I like the automatic parallelism and balancing of the batch
>> >> > scanner, but I'm looking to maintain server-side state in my iterators
>> >> over
>> >> > long-running scans. I would also like to scan over multiple tables
>> >> > concurrently. Has anyone tried hacking something together with a pool
>> >> > non-batch scanners?
>> >> >
>> >> > Adam
>> >> >
Adam Fuchs 2013-04-15, 22:19
Dave Marion 2013-04-15, 22:37
Adam Fuchs 2013-04-15, 23:01
Keith Turner 2013-04-16, 13:29
Adam Fuchs 2013-04-16, 16:33
Keith Turner 2013-04-15, 20:48