Thread: waiting for harddisk

waiting for harddisk

From
"petchimuthu lingam"
Date:
i am using postgresql 8.1.8,

Following configurations:
           shared_buffers = 5000
            work_mem = 65536
            maintenance_work_mem = 65536
            effective_cache_size = 16000
            random_page_cost = 0.1

The cpu is waiting percentage goes upto 50%, and query result comes later,

i am using normal select query ( select * from table_name ).

table has more then 6 million records.



--
With Best Regards,
Petchimuthulingam S

Re: waiting for harddisk

From
PFC
Date:
> i am using postgresql 8.1.8,
>
> Following configurations:
>            shared_buffers = 5000
>             work_mem = 65536
>             maintenance_work_mem = 65536
>             effective_cache_size = 16000
>             random_page_cost = 0.1
>
> The cpu is waiting percentage goes upto 50%, and query result comes
> later,
>
> i am using normal select query ( select * from table_name ).
>
> table has more then 6 million records.



    When you mean SELECT *, are you selecting the WHOLE 6 million records ?
Without WHERE ? Or just a few rows ?
    Please post EXPLAIN ANALYZE of your query.

Re: waiting for harddisk

From
"Scott Marlowe"
Date:
On Mon, Mar 24, 2008 at 7:05 AM, petchimuthu lingam <spmlingam@gmail.com> wrote:
> i am using postgresql 8.1.8,
>
> Following configurations:
>            shared_buffers = 5000
>             work_mem = 65536
>             maintenance_work_mem = 65536
>             effective_cache_size = 16000
>              random_page_cost = 0.1

That number, 0.1 is not logical.  anything below 1.0 is generally a
bad idea, and means that you've got some other setting wrong.

> The cpu is waiting percentage goes upto 50%, and query result comes later,
>
> i am using normal select query ( select * from table_name ).
>
> table has more then 6 million records.

You need faster disks if you want sequential scans to go faster.  Look
into a decent RAID controller (Areca, Escalade (forgot what they're
called now) or LSI) with battery backed cache.  Run RAID-10 on it with
as many drives as you can afford to throw at the problem.