postgres-xc-general Mailing List for Postgres-XC

On 9 November 2012 02:34, Koichi Suzuki <koi...@gm...> wrote:
> The result looks very similar to the one I've seen.   Maybe this is
> what we should expect with DRBD.

DRBD is very latency sensitive if not using the async modes (which
provide less data guarantees).
With SSDs and Infiniband (DRBD supports Sockets Direct Protocol over
RDMA) you can achieve close to native performance due to very low
latency of interconnect.

However, application level replication is superior in everyway.

> ----------
> Koichi Suzuki
>
>
> 2012/11/9 Vladimir Stavrinov <vst...@gm...>:
>> On Tue, Nov 6, 2012 at 4:55 AM, Koichi Suzuki <koi...@gm...> wrote:
>>
>>> Does anybody has performance data for DRBD with PostgreSQL?   The
>>> following reports that the throughput will become almost half.
>>>
>>> http://archives.postgresql.org/pgsql-performance/2007-09/msg00118.php
>>
>> I've reproduced this test against postgresql-9.1 with similar or even
>> worse results. Though experiment was not clean because of outsider
>> load on system, but kernel and drbd are not most recent. This is
>> confirmed by my other experiment with dd writes where 45% in speed was
>> lost on drbd. But this is one more argument in favor of internal HA
>> where we can expect better results.
>
> ------------------------------------------------------------------------------
> Everyone hates slow websites. So do we.
> Make your web apps faster with AppDynamics
> Download AppDynamics Lite for free today:
> http://p.sf.net/sfu/appdyn_d2d_nov
> _______________________________________________
> Postgres-xc-general mailing list
> Pos...@li...
> https://lists.sourceforge.net/lists/listinfo/postgres-xc-general



-- 
CTO | Orion Virtualisation Solutions | www.orionvm.com.au
Phone: 1300 56 99 52 | Mobile: 0428 754 846

The result looks very similar to the one I've seen.   Maybe this is
what we should expect with DRBD.
----------
Koichi Suzuki


2012/11/9 Vladimir Stavrinov <vst...@gm...>:
> On Tue, Nov 6, 2012 at 4:55 AM, Koichi Suzuki <koi...@gm...> wrote:
>
>> Does anybody has performance data for DRBD with PostgreSQL?   The
>> following reports that the throughput will become almost half.
>>
>> http://archives.postgresql.org/pgsql-performance/2007-09/msg00118.php
>
> I've reproduced this test against postgresql-9.1 with similar or even
> worse results. Though experiment was not clean because of outsider
> load on system, but kernel and drbd are not most recent. This is
> confirmed by my other experiment with dd writes where 45% in speed was
> lost on drbd. But this is one more argument in favor of internal HA
> where we can expect better results.

On Tue, Nov 6, 2012 at 4:55 AM, Koichi Suzuki <koi...@gm...> wrote:

> Does anybody has performance data for DRBD with PostgreSQL?   The
> following reports that the throughput will become almost half.
>
> http://archives.postgresql.org/pgsql-performance/2007-09/msg00118.php

I've reproduced this test against postgresql-9.1 with similar or even
worse results. Though experiment was not clean because of outsider
load on system, but kernel and drbd are not most recent. This is
confirmed by my other experiment with dd writes where 45% in speed was
lost on drbd. But this is one more argument in favor of internal HA
where we can expect better results.

On Tue, Nov 6, 2012 at 4:55 AM, Koichi Suzuki <koi...@gm...> wrote:

> Does anybody has performance data for DRBD with PostgreSQL?   The
> following reports that the throughput will become almost half.

It is interesting to compare with standby replication solution.