Re: Shared Buffers
От | Thalis A. Kalfigopoulos |
---|---|
Тема | Re: Shared Buffers |
Дата | |
Msg-id | Pine.LNX.4.21.0106212025370.27771-100000@aluminum.cs.pitt.edu обсуждение исходный текст |
Ответ на | Shared Buffers (Dennis <dennis@zserve.com>) |
Список | pgsql-general |
I had the same quetions more or less. I have 256Mb ram and set it according to the 1/4 rule to ~64M (shared_buffers=7500)after raising the /proc/sys/kernel/shmall and shmmax to 64M Today I decided to go for double the money, and switched to 128M shmem and restarted postgres with shared_buffers=15000 (8192*15000=~120M).I didn't notice any difference :) I assume that by "shared buffers" Pg means the area where common data is stored and by common data it means "index blocks"(can't think what else can be shared between processes/queries). So unless you get many users hitting different tables,you won't notice a difference. Is there a shred of truth in this or am I...? cheers, thalis On Thu, 21 Jun 2001, Dennis wrote: > > When starting postmaster, you can add the -B option to indicate the amount > of shared buffers you want postgres to use. You can also specify the value > in postgresql.conf. > > I read in the documentation that 1 shared buffer is 8192 bytes. Somone > posted a while back that it was most optimal to let postgres have about 1/4 > of the memory on a dedicated machine. > > So, I have 512 megs on a dedicated machine. I divide that by 4 and then > times the number of bytes in a meg and divide by 8192 and I get 16384. > By the 1/4 standard, it seems I should set "-B 16384" at postgres startup. > > Postgres defaults to 64 Shared Buffers. That is a big difference from > 16384. Is there any reason to my logic here or am I way off? > > Can anyone supply some more information or maybe suggest the most > optimum way to use Shared Buffers. Is the 1/4 standard good or should > I use even more.. or much less? > > Thanks in advance. > Dennis > > ---------------------------(end of broadcast)--------------------------- > TIP 3: if posting/reading through Usenet, please send an appropriate > subscribe-nomail command to majordomo@postgresql.org so that your > message can get through to the mailing list cleanly >
В списке pgsql-general по дате отправления: