>>> On Sat, 8 Oct 2005 12:09:53 +0200 (CEST), "Christian"
>>> <[EMAIL PROTECTED]> said:

[ ... ]

>> Speaking even more euphemistically the ''generic'' benchmark
>> (if that term can be stretched to cover the activities you
>> performed)

evilninja> yes, one could argue if "dd" and "tar -x.." is a
evilninja> "benchmark" at all. but i use it

Well, I use that stuff as well (but I use more the word «test»
to indicate the partial nature of such measurement, rather than
the rather more definite «benchmark»), consider for example this
note:

  http://WWW.sabi.co.UK/Notes/anno05-3rd.html#050908

and those around it, and the somewhat careful description of the
context (PC, elevator, disk types, size of the data and of the
partition, lots more...), of the measurement method and its
many limits, and other relevant details.

An example of the limits is that in my ''tests'' I used
partitions and datasets of only several GiB, which is somewhat
atypical nowadays, and I did so mostly to shorten the times
taken for testing; and I wonder how much the numbers I get apply
to partitions of several hundred GiB, which are becoming pretty
common even for home users.

evilninja> and it's a comparison at least.

Yes, but then the question is ''of what?''. For example what do
you think is being compared in the ''generic'' benchmark?

Consider for example the relative size of the kernel source tree
and that of your PC's memory, and that 'sync' does not have yet
magical :-) powers.

>> seems to me to be based on a rather original and imaginative
>> approach (and overabundant faith in the magical powers of
>> 'sync'), and its intepretation might be a challenge

evilninja> are there any "certified" generic benchmarks out
evilninja> there?

Well, not really, because the so called ''methodology'' of
running the benchmark matters _enormously_.

The result arises out of the combination of benchmark, the way it
is run, and the system. Doing precise benchmarks is impossible;
even doing just somewhat/partially useful ones is quite difficult.

Also, different usage patterns require very different benchmarks;
consider viewing some of the links in my references note:

  http://WWW.sabi.co.UK/Notes/anno05-3rd.html#050911

Warning: some of those references are not as good as others, to
speak euphemistically once more.



-------------------------------------------------------
This SF.Net email is sponsored by:
Power Architecture Resource Center: Free content, downloads, discussions,
and more. http://solutions.newsforge.com/ibmarch.tmpl
_______________________________________________
Jfs-discussion mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/jfs-discussion

Reply via email to