On Tue, Apr 26, 2011 at 9:49 AM, Claudio Freire <klaussfreire@gmail.com> wrote:
> On Tue, Apr 26, 2011 at 7:30 AM, Robert Haas <robertmhaas@gmail.com> wrote:
>> On Apr 14, 2011, at 2:49 AM, Claudio Freire <klaussfreire@gmail.com> wrote:
>>> This particular factor is not about an abstract and opaque "Workload"
>>> the server can't know about. It's about cache hit rate, and the server
>>> can indeed measure that.
>>
>> The server can and does measure hit rates for the PG buffer pool, but to my knowledge there is no clear-cut way for
PGto know whether read() is satisfied from the OS cache or a drive cache or the platter.
>
> Isn't latency an indicator?
>
> If you plot latencies, you should see three markedly obvious clusters:
> OS cache (microseconds), Drive cache (slightly slower), platter
> (tail).
What if the user is using an SSD or ramdisk?
Admittedly, in many cases, we could probably get somewhat useful
numbers this way. But I think it would be pretty expensive.
gettimeofday() is one of the reasons why running EXPLAIN ANALYZE on a
query is significantly slower than just running it normally. I bet if
we put such calls around every read() and write(), it would cause a
BIG slowdown for workloads that don't fit in shared_buffers.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
