|
|
|
Research
- Network Calculus (NetCal)
Network Calculus
(NetCal) is a collection of mathematical tools based on Min-Plus algebra,
which applies to deterministic queuing systems found in communication
networks. It is a relatively new field, with the earliest
paper on the topic dates are from 1991. There
are only two books on the subject, both published after
2000.
It can be used for example to understand and/or develop:
1. the computations for delays used in the IETF guaranteedservice;
2. why re-shaping delays can be ignored in shapers or
spacer-controllers;
3. a common model for schedulers;
4. deterministic effective bandwidth.
Netcal has two key advantages:
1. It is a method of calculation, not simulation. Hence the results are obtainable
(once the equations can be written down) with much less effort. This avoids
dealing with simulation software.
2. Netcal is based on determining a guaranteed bound on worst performance.Although
it has been used successfully in understanding the performance of a network
as a whole, it has not yet been applied to small scale systems, such as within
router. The reasons for this are threefold:
2.1. The tools needed to understand the details
of how to model a router are not yet completely available. For
example, there is not currently a systematic way of describing in network
calculus terms two processors in tandem with manufacturing blocking between
them. This kind of description is necessary in order to model a series of
processors handling a packet in sequence.
2.2. The tools that are available are new,
and appear to be very abstract. To justify a practitioner's investment
of time and effort to understand these tools, a 'proof of concept' is needed.
2.3. The network calculus provides a worst
case bound, that may be overly pessimistic in some cases. This implies
either that:
i. there is more work needed to be done on the
network calculus itself to tighten the bounds obtained by its techniques;
OR
ii. the 'worst case' specified is in fact overly pessimistic.
|
|