[orca-users] Tape Throughput Data

[Liston Bias]

On Wed, 18 Dec 2002, Blair Zajac wrote:

> Liston Bias wrote:
> >
> > First of all, commenting out the following in orcallator.se file seem to
> > work for adding tape throughput data to the combined disk throughput
> > graph.
> >
> >  if (RAW_disk[i].short_name[1] == 't' && RAW_disk[i].short_name[0] == 's') {
> >   continue;
> >  }
> >
> > Secondly, the change also creates run percentage data which you can graph
> > for the new disk (form disk_runp_st#) by making the following change to
> > to your orcallator.cfg file on orca.
> >
> > < data                  disk_runp_((?:c\d+t\d+d\d+)|(?:[ms]d\d+)|(?:st\d+))
> > < data                  disk_runp_((?:c\d+t\d+d\d+)|(?:c\d+d\d+)|(?:[ms]d\d+)|(?:st\d+))
> > ---
> > > data                  disk_runp_((?:c\d+t\d+d\d+)|(?:[ms]d\d+))
> > > data                  disk_runp_((?:c\d+t\d+d\d+)|(?:c\d+d\d+)|(?:[ms]d\d+)
> >
> > I can't see making the above change for all as many may not want the tape
> > throughput adding into disk data...
> >
> > Since my goal of this was to find out my write throughput to tape on
> > backup server and 99% of my write throughput is tape, the combined disk
> > throughput graph seems to work well for me.
> >
> > However, I was wonder if anyone has ever done coding on getting throughput
> > graphs for the individual disk on system?  I imagine it would require some
> > combination of the total throughput and individual runp data to generate
> > new graphs.
> >
> > Because of the "ridiculously" large number of runp graphs generated on our
> > system with hundreds of disks, I usually turn of runp.  Has anyone done
> > work on using same cfg file to graphs different data for different
> > systems?
> >
> > I have no immediate need for either of above, but would be nice to have.
>
> Would there be any need for a total disk and tape throughput plot?
> Right now there orcallator.se and orcallator.cfg measure separate
> measurements for the total disk and total tape throughput and plot
> these measurements separately.
>
> Would the combined disk and tape be useful to see the total throughput
> that the system is managing?

I could see combining these into single graph so you can better compare
but I think it will vary what people are looking to compare.  I can
envision no reason why a "total" (tape + disk) graph would be useful.  We
actually use the throughput charts to find correlation between tape
throughput and network throughput.

- Liston