iiit-ih@pyramid.swansea.ac.UK (Ivan Izikowitz) (04/28/89)
I am currently involved in a British Standards exercise looking at
Time Critical Communication Architectures (TCCA) in the context of
the OSI reference model. We are currently examining metrics for network
performance so that users may have some tools to use in the choice of
or configuration of time critical networks.
We find the current means of quantifying performance (ie. octects per
second through layers etc.) to be rather unsuitable for our purposes. I
have seen some discussion on the list about the optimization of ASN.1
encoding and the metrics used for that purpose. My questions are...
- Have any metrics for the performance of layered protocols been
defined in the time domain;
- Are any other groups addressing this problem;
We are interested in any input in this area and are particularly
concerned that we do not replicate any effort already expended in
this area.
Any assistance/information would be most welcome.
Ivan
##############################################################################
# Ivan Izikowitz, Institute for Industrial Information Technology, #
# Innovation Centre, Swansea SA2 8PP #
# Phone: (+44) 792 295213 # JANET: iiit-ih@uk.ac.swan.pyr #
# Fax: (+44) 792 295532 # UUCP: ..!ukc!cybaswan.UUCP!iiit-ih #
##############################################################################
My views are probably my own! rhc@HPLB.HPL.HP.COM (Robert Cole) (05/04/89)
> I am currently involved in a British Standards exercise looking at > Time Critical Communication Architectures (TCCA) in the context of > the OSI reference model. We are currently examining metrics for network > performance so that users may have some tools to use in the choice of > or configuration of time critical networks. > We find the current means of quantifying performance (ie. octects per > second through layers etc.) to be rather unsuitable for our purposes. I > have seen some discussion on the list about the optimization of ASN.1 > encoding and the metrics used for that purpose. My questions are... > - Have any metrics for the performance of layered protocols been > defined in the time domain; > - Are any other groups addressing this problem; > We are interested in any input in this area and are particularly > concerned that we do not replicate any effort already expended in > this area. > Any assistance/information would be most welcome. Ivan, I am interested in your project. I am currently thinking about the problems of moving a time synchronisation protocol over to OSI. The protocol runs on the DARPA Internet where it uses the datagram service. The protocol relies on the round trip time being fairly constant. That is, only delays due to queueing should afect the round trip time. When moving over to OSI it seems that the connection oriented protocols at each layer are not only going to interfere with each other, but any retransmissions due to lost packets at lower layers will cause severe disrupton to the round trip time varience. Is your project looking at this type of application? Robert.
iiit-ih@pyramid.swansea.ac.UK (Ivan Izikowitz) (05/05/89)
Robert,
It is interesting to hear of the work that you are involved in. The
BSI group is looking into ways of quantifying the performance of OSI
protocols/profiles so that implementors like yourself will be able to
get some sort of guide as to the bounds of the system in the temporal domain.
These metrics would possibly be used by vendors to specify the performance
of their products and consequently enable users to compare similar products
from different vendors. As you know, in the connection-oriented approach,
multiple associations are active at any one time - each requiring a share of
the available resources. The net result of such a system is highly non-
deterministic. As a bit of background....
The working group at the BSI is an initiative brought about by the poor
performance of ODISI networks in time-critical environments (I'm sure you
have heard all the horror stories about MAP!!!) We are looking at the
use of current OSAII standards in time-critical environments and if need be,
we will make recommendations for new standards and/or additions to existing
standards. We are currently looking at the "Quality of Service" feature
provided in the transport layer (in conjunction with the session layer) as
a possible means of improving the response time.
As you may appreciate I have only skimmed the surface of such issues but
if you have any definite questions about the work I will be happy to reply.
Aside from this, in my official capacity, we are involved in time
synchronization over networks (including distributed clock synchronization)
and I would be interested in hearing about your work (if it is not commercially
sensitive of course).
Regards,
Ivan
PS. If you are indeed actively involved in time-critical comms over OSI
and you feel that the group may benefit from your experience AND there
is a budget at HP for standards work, I could possibly look into nominating
you as a member of the group (we meet in London approx. every 4-5 weeks).