Dear Colleagues, I was just wondering if benchmarks on the performance of
the srsbuild process were available or were foreseen to be exploited. I
agree that since the srsbuild is relying on the icarus scripts which can
be changed at any site appropriately, at least theoretically it could be
troublesome to compare indexing performance. Moreover, since the
indexing time is certainly dependent upon the size and the format of the
database used there are various parameters to be taken into
consideration. Certainly wether the data and index disks are local or
remotely mounted that as well can make a difference .. as well as the
machine load etc... Finally, since the index generation process is
composed of several steps, the choice of which of the srsbuild atomic
operations is also a variable.
However, I think that is useful know quantitatively about the performance
of the srsbuild step and its monitoring on wide range of platforms.
Hereafter I provide an example of a record which one may consider as
prototype for this kind of data-collection ... restricting to the very
first srsbuild command launched, the one which is launched with the
parameters: srsbuild XXXXX -xdir XXX: -odir XXX: -s unix -nn
---------------------------
A) Hardware parameters
Processor type:
CPU speed:
data location:
index location:
B) Operating System information
Operating System Name:
Operating System Version:
C) Database specific information
Database:
Format:
DBRelease:
SRSRelease:
D) Performance
real cpu time:
user cpu time:
sys cpu time:
------------------------
Since certain Organizations could feel that this data contain
proprietary information, the report could be then "anonymized" by the SRS
maintainers, which each of us trust fully.
This is just a suggestion, which hope will stimulate discussion and
actions on this important aspect, which fits into the general area of
performance in biocomputing...
regards and compliments
Dr. Luca I.G. TOLDO
-------------------==== Posted via Deja News ====-----------------------
http://www.dejanews.com/ Search, Read, Post to Usenet