Re: Compression of full-page-writes

On 11/10/13 19:06, Andres Freund wrote:
> On 2013-10-11 09:22:50 +0530, Amit Kapila wrote:
>> I think it will be difficult to prove by using any compression
>> algorithm, that it compresses in most of the scenario's.
>> In many cases it can so happen that the WAL will also not be reduced
>> and tps can also come down if the data is non-compressible, because
>> any compression algorithm will have to try to compress the data and it
>> will burn some cpu for that, which inturn will reduce tps.
> Then those concepts maybe aren't such a good idea after all. Storing
> lots of compressible data in an uncompressed fashion isn't an all that
> common usecase. I most certainly don't want postgres to optimize for
> blank padded data, especially if it can hurt other scenarios. Just not
> enough benefit.
> That said, I actually have relatively high hopes for compressing full
> page writes. There often enough is lot of repetitiveness between rows on
> the same page that it should be useful outside of such strange
> scenarios. But maybe pglz is just not a good fit for this, it really
> isn't a very good algorithm in this day and aage.
>
Hm,. There is a clear benefit for compressible data and clearly
no benefit from incompressible data..

how about letting autovacuum "taste" the compressibillity of
pages on per relation/index basis and set a flag that triggers
this functionality where it provides a benefit?

not hugely more magical than figuring out wether the data ends up
in the heap or in a toast table as it is now.

--
Jesper