Manfred Koizar wrote:
[ still brainstorming ... ]
On Thu, 21 Aug 2003 17:16:50 -0400, Tom Lane [EMAIL PROTECTED]
wrote:
Whenever a backend encounters a dead tuple it inserts a reference to
its page into the RSM.
This assumes that backends will visit dead tuples with significant
[EMAIL PROTECTED] (Tom Lane) wrote in message news:[EMAIL PROTECTED]...
Manfred Koizar [EMAIL PROTECTED] writes:
better. AFAICS Vivek's problem is that it is hard enough to hold a
good part of the working set in the cache, and still his disks are
saturated. Now a VACUUM not only adds one
[EMAIL PROTECTED] (Grant Succeeded) writes:
The best for me by far, is to get the OS to *not* cache stuff. As
long as the database uses the information it inherently has available,
it can make far more effective use of the same amount of memory the OS
would have used to cache the whole
On Thu, 21 Aug 2003, Tom Lane wrote:
We have had some people looking at improved buffer management
algorithms; LRU-2 or something smarter would help. I dunno whether
we can dissuade the kernel from flushing its cache though.
Using open/read/write system calls, you can't. You can always use
Manfred Koizar [EMAIL PROTECTED] writes:
better. AFAICS Vivek's problem is that it is hard enough to hold a
good part of the working set in the cache, and still his disks are
saturated. Now a VACUUM not only adds one more process to disk I/O
contention, but also makes sure that the working
Manfred Koizar [EMAIL PROTECTED] writes:
Good point. What about: Whenever a backend *deletes* a tuple it
inserts a reference to its page into the RSM? Then an entry in the
RSM doesn't necessarily mean that the referenced page has reclaimable
space, but it would still be valueable
On Wed, 20 Aug 2003 15:39:26 -0400, Tom Lane [EMAIL PROTECTED]
wrote:
But I think the real point here is that there's no reason to think that
doing tuple deletion on-the-fly in foreground transactions is superior
to doing it in background with a vacuum process. You're taking what
should be
[ still brainstorming ... ]
On Thu, 21 Aug 2003 17:16:50 -0400, Tom Lane [EMAIL PROTECTED]
wrote:
Whenever a backend encounters a dead tuple it inserts a reference to
its page into the RSM.
This assumes that backends will visit dead tuples with significant
probability. I doubt that assumption
On Thu, 21 Aug 2003 17:56:02 -0400, Tom Lane [EMAIL PROTECTED]
wrote:
Conceivably it could be a win, though,
if you could do frequent vacuum decents and only a full-scan vacuum
once in awhile (once a day maybe).
That's what I had in mind; similar to the current situation where you
can avoid