Re: WIP patch for hint bit i/o mitigation

On 07-Nov-2012, at 15:46, Amit Kapila <amit(dot)kapila(at)huawei(dot)com> wrote:

>> From: pgsql-hackers-owner(at)postgresql(dot)org [mailto:pgsql-hackers-
>> owner(at)postgresql(dot)org] On Behalf Of Merlin Moncure
>> Sent: Wednesday, November 07, 2012 5:26 AM
>> To: PostgreSQL-development
>> Cc: Atri Sharma
>> Subject: [HACKERS] WIP patch for hint bit i/o mitigation
>>
>> Following the sig is a first cut at a patch (written by Atri) that
>> attempts to mitigate hint bit i/o penalty when many pages worth of
>> tuples are sequentially written out with the same transaction id.
>> There have been other attempts to deal with this problem that fit
>> niche cases (especially those that create the table in the same
>> transaction as the one inserting) that work but don't really solve the
>> problem generally.
>>
>> I previously attacked this problem ([1], [2]) and came up with a patch
>> that cached hint bits inside tqual.c. The patch was pulled for a few
>> reasons:
>>
>> 1) a lot of complexity without proper justification
>> 2) sketchy cache replacement algorithm
>> 3) I manged to misspell 'committed' just about everywhere
>> 4) invalidation?
>>
>> Issues 1-3 could have been worked out but #4 was making me think the
>> problem was a nonstarter, or at least, 'too much too soon'. The tuple
>> visibility routines are in a very tight code path and having to deal
>> with various things in the backend that could cause the xid to become
>> stale were making me nervous. A smaller, simpler patch might be the
>> ticket.
>
> About invalidation, I think the cached xid can become invalid due to xid
> wraparound.
> So for that one way could be to invalidate it through Vacuum.
>
> Though I am not sure what all other things can make cached id as invalid,
> but I think once we
> can think what other ways can make cached id invalid, then we can see if
> there is a solution to address
> them.
>
>
> With Regards,
> Amit Kapila.
>
>
>
> --
> Sent via pgsql-hackers mailing list (pgsql-hackers(at)postgresql(dot)org)
> To make changes to your subscription:
> http://www.postgresql.org/mailpref/pgsql-hackers

As we are now dealing with only the last xid(please refer to the details of the patch attached), the invalidation issues are not significant any more.

Regards,

Atri