Re: Database block lifecycle

Ok,  I wasn't precisely enough, you are right. It's brand new server,
nothing is yet configured and we have not even os installed. The number
was the overall count we expect for a whole cluster.

But the main question is: is it possible to completely avoid disk read
if there is huge amount of RAM available?


Am 13.08.2014 00:39, schrieb John R Pierce:
> On 8/12/2014 3:29 PM, pinker wrote:
>> yes, I know the count is quite high. It is the max value we've
>> estimated, but probably on average day it will be 100-200, and yes we
>> use pgpool.
>
>
> if you're using a pooler, then why would you be using 200 concurrent
> connections, unless you have a 50 or 100 CPU cores/threads ?
>
> if you have 1000 transactions to execute on a 32 core server, and you
> try and do 200 at once, it will take longer than if you do 64 at a
> time and let the rest queue up.
>