[Orca-users] Re: Digest Number 213

[Blair Zajac]

The only way to verify the data you're seeing is to run through the
web server access logs in 5 seconds intervals and see if you get
the same large number of hits per second that orcallator is
measuring.  A simple Perl script would probably do the job here.

Which version of orcallator.se are you using?  Later versions had
some additional code to make sure that the measurements took into
account that a busy CPU could mean that the actual measured interval
is not exactly 5 seconds.

Regards,
Blair

paul beard wrote:
> 
> I'm attaching a plot that shows what I'm talking about.
> 
> I also have a sampler of my own that I use, and it's accurate over the
> longterm: I don't think we see any real spikes.
> 
> Using sample size of 1000.
> For logfile /logs/fizzylab.access:
>         18 hits/second, 1090 hits/minute, 65454 hits/hour, 1570909
> hits/day
> 
> orca-users at egroups.com wrote:
> 
> >
> > Message: 6
> >    Date: Mon, 15 Jan 2001 11:06:06 -0800
> >    From: paul beard <paul at fizzylab.com>
> > Subject: httpops/second plotting bug??
> >
> > It's entirely possible I don't understand how the peak 5 second plots
> > are generated vs the average values, but I know the average values are
> > right and want to be able to defend, if possible, the peak values. I
> > don't want anyone making assumptions about capacity on those values. Is
> > this a bug (I note that the sample plots in the documentation don't show
> > near as much variation).
> >
> > For instance, where now I see 9 in my 5 minute average, my 5 second
> > average is 64. I know the 9 is right.
> >
> > Can someone 'splain this?
> >
> > --
> > Paul Beard / 206 728 4282 x138
> > FizzyLab / 2025 First Ave, Suite 1150
> > Seattle WA 98121
> >
> > Q:      What is the difference between Texas and yogurt?
> > A:      Yogurt has culture.
> >
> > ________________________________________________________________________
> > ________________________________________________________________________
> >
> > Message: 7
> >    Date: Mon, 15 Jan 2001 12:04:36 -0800
> >    From: Blair Zajac <blair at akamai.com>
> > Subject: Re: httpops/second plotting bug??
> >
> > You're talking about the web server hits per second?  When orcallator
> > is monitoring the web server, it sleeps for 5 seconds, counts the
> > number of new lines in the web server access log, goes back to sleep
> > for another 5 seconds, and repeats this.  After a 5 minute interval,
> > the average hits per second is calculated for the whole 5 minute
> > interval.  Over the 5 minute interval, the largest number of hits in a
> > 5 second interval is saved and divided by 5 seconds to get the peak
> > hits per second.  A difference between 9 and 64 does sound high, but
> > you can check the access log to see if you get variations in hits
> > like this.
> >
> > Regards,
> > Blair
> >
> > paul beard wrote:
> > >
> > > It's entirely possible I don't understand how the peak 5 second plots
> > > are generated vs the average values, but I know the average values are
> > > right and want to be able to defend, if possible, the peak values. I
> > > don't want anyone making assumptions about capacity on those values. Is
> > > this a bug (I note that the sample plots in the documentation don't show
> > > near as much variation).
> > >
> > > For instance, where now I see 9 in my 5 minute average, my 5 second
> > > average is 64. I know the 9 is right.
> > >
> > > Can someone 'splain this?
> > >
> > > --
> > > Paul Beard / 206 728 4282 x138
> > > FizzyLab / 2025 First Ave, Suite 1150
> > > Seattle WA 98121
> > >
> > > Q:      What is the difference between Texas and yogurt?
> > > A:      Yogurt has culture.
> >
> > ________________________________________________________________________
> > ________________________________________________________________________
> 
> --
> Paul Beard / 206 728 4282 x138
> FizzyLab / 2025 First Ave, Suite 1150
> Seattle WA 98121
> 
> Nitwit ideas are for emergencies.  You use them when you've got nothing
> else to try.  If they work, they go in the Book.  Otherwise you follow
> the Book, which is largely a collection of nitwit ideas that worked.
>                 -- Larry Niven, "The Mote in God's Eye"
> 
>   ------------------------------------------------------------------------
>  [Image]