On Thursday 15 Dec 2005 15:50, brian davies wrote:
> Are people going/am stating bandwidth rates in th ewili by using
> network monitoring tools or calculating effective bandwidth for
> service challenge throughput tests? by effective bandwidth i mean
> numberoffiles*filesize/totaltimeforalltransfers
> I would be interested to know both if people have the capacity to
> calculate both.
> brian
Hi Brian
So far I have been using a combination of both, at least using Ganglia to see
the actual rate changing and spot peaks and breaks.
However, what really matters (and should be reported in the wiki) is the
aggregate rate, which should be calculated as
(nFiles * size)/time
If you use my python script[1], in addition to giving you the aggregate rate,
it will log the status changes of all the individual files in the transfer,
so once I write the object method (v0.2?), it will be able to calculate a
reasonable bandwidth average over about 2*(single file transfer time). (Which
will be more important to do for transfers that break or suffer some file
loss/retry.)
In the wiki, as you've probably seen, Chris Brew had the nifty idea of making
a link from the transfer test table to his local site page, where he puts
down transfer details and notes. Grieg's been doing this too, and I'd
encourage other sites to follow this method.
Cheers
Graeme
[1] http://www.physics.gla.ac.uk/~graeme/scripts/
--
--------------------------------------------------------------------
Dr Graeme Stewart http://www.physics.gla.ac.uk/~graeme/
GridPP DM Wiki http://wiki.gridpp.ac.uk/wiki/Data_Management
|