usleepless@gmail.com writes:
> Hi List and Chris,
>
> On 11/10/06, Chris Browne <cbbrowne@acm.org> wrote:
>> The defaults for various interesting things were...
>> max_connections = 100
>> shared_buffers = 32MB
>> max_fsm_pages = 204800
>>
>> The default work_mem was 1MB, which is arguably low, on a machine with
>> 15GB of RAM :-).
>>
>> But frankly, I don't see any particular problem with the defaults.
>> 32MB is *not* an over-tiny amount of shared memory; it's enough to be
>> meaningful for more than merely a trivial workload.
>
> "trivial", that's the point. there are people who are loading huge
> tables into postgresql the same way they do into mysql. and then they
> compare some queries. and then go back to mysql.
>
> i think we need a make tune command which can configure at least up to
> 50% of the machine's physical hardware.
I disagree.
With the improvements in the quality of buffer usage, if the buffer
sizing has improved by a factor of 64, I do NOT see there being a
similar kind of improvement to be had by trying to get exceedingly
automated about going further.
To *in theory* get another 64-fold improvement requires that people be
prepared to devote 2GB of RAM to shared memory, which seems very
unreasonable.
The fact is that the changes that have been made are, *RIGHT NOW*, an
enormous improvement. Increasing buffer sizes further will NOT have
the same "return on investment."
--
let name="cbbrowne" and tld="linuxdatabases.info" in name ^ "@" ^ tld;;
http://www3.sympatico.ca/cbbrowne/linux.html
'There is no substitute for good manners, except, perhaps, fast
reflexes.' -- random unix fortune
