Re: Utilization affect on throughput [7:76157] posted 09/28/2003
- Subject: Re: Utilization affect on throughput [7:76157]
- From: "Zsombor Papp" <nobody@xxxxxxxxxxxxxx>
- Date: Sun, 28 Sep 2003 22:35:20 GMT
It all depends on how you define these terms. Sphon's definition is that
"throughput is a measure of how much data can be passed across a medium in a
stated period of time." I think I understand this, and it really does imply
that delay will affect throughput. I just think it's a useless measure.
However what I am really curious about is goodput. The freakin' CGI
interface ate half of my previous post due to reckless usage of less-than
characters (talk about lack of goodput! :), so here it is again:
I assume "goodput is less than or equal to throughput" must hold true in any
case, so either there are factors that account for the "goodput is less than
throughput" case, or these two terms are equivalent.
Small example (that might also help understand the difference between
capacity and throughput): assume a system that can transmit 100 bps
(capacity) from A to B so that each bit spends 1 second (delay) within the
system. If I send 100 bits from A to B and all of them arrive as they
should, then throughput is 50bps according to Sphon's definition, as it took
2 seconds to transmit 100 bits. (Notice how the throughput will increase to
66, 75, 80, ... bps if I send 200, 300, 400, ... bits; just a small hint why
I think "throughput" is a useless term. :)
What would goodput be in the above case? I assume its "less than or equal
to" 50bps, but can it be exactly 50bps?
Now assume that one of the 100 bits changed during transmission due to a
sunspot, or whatever. I guess it's clear that capacity is still 100 bps
(although you could argue that capacity is really 100bps-BER). But what
would throughput and goodput be?
Can we say that throughput is still 50bps, but goodput is less than 50bps,
and possibly zero, depending on the error recovery used by a higher layer?
Marc Thach Xuan Ky wrote:
> I still confuse throughput and capacity with respect to delay.
> If I
> have a video streaming app, then two clients will see the same
> regardless of whether or not one is delayed by a second going
> 200,000 km of wire. The steady-state throughput is therefore
> the same
> in each case. So from the networkers viewpont, looking at the
> through a router port, delay does not affect throughput.
> Granted that a
> transaction-based application may be sensitive to delay, that's
> nature of network apps. So from an application designers point
> of view
> then the network may be lowering the throughput of the system,
> but the
> app designer is not in a position to evaluate the throughput of
> particular router along the path, so I cannot see how it is
> correct to
> say that the throughput of a router can be affected by delay.
> Marc TXK
Message Posted at:
**Please support GroupStudy by purchasing from the GroupStudy Store:
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html