incubator-accumulo-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Keith Turner <ke...@deenlo.com>
Subject Re: benchmarks
Date Mon, 13 Feb 2012 23:23:02 GMT
I would be interested in working w/ someone on this.  Right now I am
concentrating on getting 1.4 release ready, and that is keeping me
pretty busy.  After we release 1.4 I may have some time.

On Mon, Feb 13, 2012 at 5:47 PM, Joey Echeverria <joey@cloudera.com> wrote:
> I thought I heard Aaron volunteering ;)
>
> -Joey
>
> On Feb 13, 2012, at 16:58, Keith Turner <keith@deenlo.com> wrote:
>
>> I think its a good idea, but who is going to take the time to do it?
>>
>> On Mon, Feb 13, 2012 at 4:42 PM, Joey Echeverria <joey@cloudera.com> wrote:
>>> I think benchmarks are interesting and useful, but they're not the whole
>>> story. If we can collaborate with the Hypertable testing, that'd be best as
>>> it should avoid getting into a benchmark shoot out where each community
>>> feels compelled to one-up the other.
>>>
>>> -Joey
>>>
>>>
>>> On Mon, Feb 13, 2012 at 4:07 PM, Aaron Cordova <aaron@cordovas.org> wrote:
>>>>
>>>> so .. we all know that benchmarks can be misleading but ...
>>>>
>>>> There seems to be some excitement surrounding the recent Hypertable vs
>>>> Hbase benchmark, particularly because hbase (0.90) failed to complete some
>>>> tests, and because the tests seem to be based on those outlined in the
>>>> original BigTable paper.
>>>>
>>>> Hypertable has stated (on Twitter) that they're going to re-run the
>>>> benchmark on Hbase 0.92, and that they hope for participation from Hbase.
I
>>>> wondered whether it would be in the Accumulo community's interest to
>>>> participate as well?
>>>>
>>>> I believe that because Accumulo uses native management of memory for it's
>>>> ingest operations that it won't have the same problem Hbase 0.90 did with
>>>> the tests involving lots of small (10k and 100k) inserts. We know that
>>>> Accumulo stands apart somewhat because of it's unique features, but
>>>> performance could be an important differentiator too, or at least should
not
>>>> be a weakness ...
>>>>
>>>> Thoughts?
>>>>
>>>> Aaron
>>>
>>>
>>>
>>>
>>> --
>>> Joseph Echeverria
>>> Cloudera, Inc.
>>> 443.305.9434
>>>

Mime
View raw message