🚀 go-pugleaf

> Date: Fri, 28 Aug 92 13:27:32 PDT
> From: Pasha Quadri <quadri@last-call.cisco.com>
> To: exubho@exu.ericsson.se
> Cc: cisco@spot.colorado.edu
> Subject: Re: X.25 line utilization???
> 
> > 
> > Hi,
> > 
> > I have some statistics I have been gathering from one of my packet
> > switch nodes within my company's private X.25 network.  I have been
> > monitoring the incoming and outgoing bit/sec data rates between
> > Richardson Texas and Stockholm Sweden.
> > 
> > We have a full duplex communications link capable of 112
> > kilobits/sec of data throughput.  We have recently gotten into a
> > quarrel over what the utilization of the link is.  
> > 
> > Here is the scenario.  My statistics tell me that I have the
> > following:
> > 
> > 40,000 bits/sec input data rate
> > 60,000 bits/sec output data rate
> > 
> > I think that the utilization of our line during this time (expressed
> > as a percentage) is ((40,000 + 60,000)/112000) * 100%
> > 
> > while some others here think that the utilization of our line
> > (expressed as a percentage) is gotten by taking the higher of the
> > two data rate figures and dividing that by the maximum data rate, or
> > (60,000/112000)*100%.
> > 
> > Does anyone have any input on which is the right way to express the
> > utilization of our link?  Any input would be appreciated.
> > 
> > Blake Holman
> > Ericsson Network Systems, Inc.
> > Richardson, Texas
> > (214) 997-6784
> > 
> 
> 
> Hello Blake,
> 
> Since serial lines are full duplex dividing by the bandwidth of the link
> will just not work. Hence greater of either input OR output bytes, bits, or
> packets per second be taken and divided by the BW of the link.
> 
> One of our Engineers had written an article on this and a formula for
> average utilization was :
> 
>  average utilization = [(bytes-input OR bytes-output) * 8] / ifSpeed
> 
>    The OR above denotes a case where you should take the larger
> of the input and output values to divide by the bandwidth (ifSpeed) to
> determine the true link utilization.
> 
> Hope this helps.
> 
> 
> Regards
> 
> Pasha Quadri
> cisco Cust.Engg

I agree with Blake (except that I think the two rates should be averaged
instead of simply summed), and think a combination of the input and output
is appropriate.  Since the traffic in the two directions is mostly
independent, you could argue that full utilization in one direction and
zero utilization in the other (as an extreme end condition) would
represent only half of the total traffic that could be traversing the
line.  As a real world example, a large ftp session could easily
consume the bulk of the available bandwidth in one direction, but
would still allow another large ftp transfer to take place in the
other direction.  To me, saying that a line is being utilized at 100%
of capacity means that no more traffic will fit.  That is not the
case when one bases such utilization numbers on just the larger of
the IN/OUT traffic.  This may not be the most mathematical approach
to this problem, but most users (and bosses) don't care about technically
correct statistics - they want numbers that make sense, and that give
a real measure of how much of what they're paying for is actually
getting used.

Dana