From: | Allan Kamau <kamauallan(at)gmail(dot)com> |
---|---|
To: | Postgres General Postgres General <pgsql-general(at)postgresql(dot)org> |
Subject: | CPU choice for postgreSQL |
Date: | 2011-11-26 07:23:23 |
Message-ID: | CAF3N6oSu4iKoBEd1e0Y=A_4E7q6ouaZQjNJuB9sqx1UbSR2reA@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general |
Hello,
I do have a read-only table having a field having long varbit data (of
length 6000). And I have a function that performs various aggregate
bitAND and bitOR operations on this field and other fields of this
table. This function does not explicitly write any data to disk (here
is hardly any disk activity) and consumes 100% resource of a single
core during it's execution and I can open several connections and
invoke this function repeatedly on each connection but with different
parameters. Currently I have copies of the read-only table in various
physical servers and I open several connections to each one of these
servers and repeatedly execute the above function on each one of this
connections in parallel.
Now I would like to procure new hardware to compliment my efforts.
There are substantial amounts of data moving from RAM to CPU and it
seems the most important hardware items would be CPU and RAM speed.
This are my current options all with fast RAM.
1) Build 4 separate PCs based on AMD Phenom II X6 1100T CPU with
mini-ITX motherboard each.
2) Build 4 separate PCs based on Intel Core i7-2500K CPU with mini-ITX
motherboard each.
3) Build 1 server having 4 AMD 12 core Magny-cours CPUs (on one Tyan
motherboard).
4) Build 1 server having 4 AMD 16 core Interlagos CPUs (on one motherboard).
Which of these options and hopefully others should I consider.
Regards,
Allan.
From | Date | Subject | |
---|---|---|---|
Next Message | Condor | 2011-11-26 07:28:06 | Re: How to add conversion between LATIN1 and WIN1251 ? |
Previous Message | Tomas Vondra | 2011-11-26 03:42:52 | Re: General performance/load issue |