Estimate of Research Bandwidth for the EBONE Backbone
Phil Jones
3 September 1992
-------------------------------------------------------------------
0. Contents
1. Introduction
2. Background
3. Data
4. Estimates
5. Postscript
1. Introduction
The Ebone Management Committee requested that EAT provide estimates
of what the speeds of the EBONE backbone links should be to be close
to estimated requirements. This is expressed as a single figure.
2. Background
EAT members were very concerned that the background to the estimates
was clearly stated lest the estimates be misinterpreted.
Concerns have been expressed that large estimates would be
undervalued, for example they might appear unrealistic in financial
terms, or in comparison with what is available at the time of
writing (or indeed reading). This paper addresses this view by:
- drawing attention to it here;
- documenting the background and assumptions behind the estimates.
It has to be pointed out that estimates are, of course, made with
varying degrees of certainty. Some of the factors affecting the
accuracy of estimates would be:
- the effect of Parkinson's law (variant of);
- the fact that projecting traffic to the future is significantly
more difficult where the resources involved are being used at
saturation of near-saturations levels. This is because the
potential increase in usage is constrained by the lack of
available resource;
- the effect of future events (e.g. connection of additional
networks to the Ebone or upgraded (or degraded) links.
There will be loads and linkspeeds at which EBONE will be scarcely
viable, especially for interactive traffic. There are those that
think that EBONE is close to that already.
There are (as at August 1992) those who think that the arrangements
for inter-continental access to and from Ebone that are in place at
the time of writing strongly militate against good estimates of the
EBONE backbone speeds. Issues that some considers relevant here are:
- the lack of inter-continental resources overall;
- the lack of inter-continental resources belonging to the EBONE
consortium;
- the technical difficulties involved in backing up
inter-continental links;
- the difficulties involved in sharing inter-continental links
(see RFC1346);
- the fact that multiple agencies are involved in the arrangements
at both ends of almost all such links.
In making the estimates, the intention is to come with a single
figure for the whole EBONE backbone, effectively an estimates for an
EBONE pentagon where all links are at the same speed, and assuming
five EBS's installations configured in a pentagon as planned for
release as a resilient EBONE in September 1992.
3. Data (Assumed)
Sum of the speeds of RBS connections at the EBS installations as
follows:
Stockholm 3 Mbps
Amsterdam 3 Mbps
Geneva 10 Mbps
Montpellier 2.5 Mbps
London 4 Mbps
------------
22.5 Mbps
Notes:
- this takes the connections of the RBS to a given EBS via routers
on the same LAN to represent a total connectivity of 2 Mbps;
- it would be valuable to see the calculations redone based on the
bids for connectivity for Ebone93.
4. Estimates
Three estimates are provided assuming:
a) that the backbone speed should be half the total of RBS access
speeds: 11 Mbps;
b) that the backbone speed should be equal to the highest of the
total access speeds at the five EBS installations: 10 Mbps;
c) gut feeling, which is 3 Mbps. (Note that this is roughly an order
of magnitude greater that that currently available).
5. Postscript (Written Later)
My thanks to Bernhard Stockman for his help with this paper.
Apologies are due for the delay in the publication of the paper.
This was due to the pressure of other work on both Bernhard and the
author. The results were provided in a written report distributed to
the Ebone Management Committee for its meeting in September 1992,
and the background explained to the committee at the time by the the
author.
Other, much more sophisticated, analysis could be made of the
requirements on a network, e.g. by more detailed analysis of the
usage or by other means (e.g. surveys of the customers requesting
estimates of requirements). For example, consideration of the usage
of the original Ebone pentagon could have been done based on the
source and destination RBS' of the traffic. Based on these, one
could have estimated the value of:
- upgrading the five links on the original Ebone backbone;
- installing one or more of the five possible cross-connections on
the Ebone original backbone;
- installing a further EBS in any one of of number of possible
locations around Europe.
The author's view (expressed at the Ebone Management Committee and
to the EAT) is this should have been done before the decision was
taken to install a sixth EBS.