Re: Initial prefetch performance testing

Bruce Momjian wrote:
> Ron Mayer wrote:
>> Even more often on systems I see these days, "spindles"
>> is an implementation detail that the DBA has no way to know
>> what the correct value is.
>>
>> For example, on our sites hosted with Amazon's compute cloud (a great
>> place to host web sites), I know nothing about spindles, but know
>> about Amazon Elastic Block Store[2]'s and Instance Store's[1]. I
>> have some specs and are able to run benchmarks on them; but couldn't
>> guess how many spindles my X% of the N-disk device that corresponds
>> to. For another example, some of our salesguys with SSD drives
>> have 0 spindles on their demo machines.
>>
>> I'd rather a parameter that expressed things more in terms of
>> measurable quantities -- perhaps seeks/second? perhaps
>> random-access/sequential-access times?
>
> I assume SAN users might not know the number of spindles either.

Yeah. Nevertheless I like the way effective_spindle_count works, as
opposed to an unintuitive "number of blocks to prefetch" (assuming the
formula we use to turn the former into latter works). Perhaps we should
keep the meaning the same, but call it "effective_io_concurrency"?
Something that conveys the idea of "how many simultaneous I/O requests
the I/O subsystem can handle", without referring to any specific
technology. That concept applies to SANs and RAM drives as well.

--
Heikki Linnakangas
EnterpriseDB http://www.enterprisedb.com