|
Khaled,
The
ambiguous part still lies in the conversation path. The hub node can sink 200%
and source 200% because it is connected to 2 rings.
I'm
not adding both transmits and receive into the traffic load number.
Whether the hub can process all the traffic (200% each
direction) is outside the scope of the MAC.
e.g.
The hub can source 200% on both rings, each talk to 1/2 the station. This is a
spatial reuse property.
Proposal:
In the
hub application:
There
are 16 stations, 1 to 16 like a hex clock. Station 16 is the Hub.
Stations 1 to 15 all send traffic to station 16 on the same ring e.g. in
the clockwise ring
Sum of all station traffic is 200%
of link (10G) capacity all turned on at time T1.
Traffic source from each station is
1.3G each, destination is the hub.
Regards,
Harry
Harry,
Thanks for pointing out this ambiguity.
We had discussed that overcapacity would mean
200%. This should be added to the chart.
Now, the question becomes if that would mean that
the hub generates 100% capacity and receives 100%?
I suggest that the answer is NO.
The reason is that this would not be over
capacity since we're using dual rings. This would only be considered over
capacity if we're using one ring.
So, for dual rings (as in the charts), we should
have the hub generate 200% the ring capacity (i.e. 20Gbps is running a 10Gbps
ring), and also receive 200% of the ring capacity generated by the other
nodes.
Does this help clarify the
ambiguity?
Any discussions or suggestions?
Khaled Amer President, AmerNet Inc. Architecture Analysis and
Performance Modeling Specialists Address: 13711
Solitaire Way, Irvine, CA
92620 Phone:
(949)552-1114
Fax:
(949)552-1116
e-mail: khaledamer@xxxxxxx
Web: www.performancemodeling.com
----- Original Message -----
Sent: Friday, April 20, 2001 6:24
AM
Subject: RE: [RPRWG] RPR Perf: Metrics
for Phase I
Khaled:
I'm not clear as to the conversations (source to
destination) profile on the ring as outlined on page 22 of the attached
document.
How is the ring over capacity?
and what is the overcapacity number? 150%,
200%...
Regards,
Harry
All,
I've been getting questions about Phase I of the performance modeling
and what's included.
Throughout the work of the ad-hoc group we came to an agreement to
these metrics and parameters.
We've been trying to put it in a document, but it got into too much
details related to headers, signals and interfaces which hindered its
progress.
In my opinion, based on several years of performance modeling as a
career, all what is needed is in the attached presentation. There are some
minor points that are left open which provide us with some flexibility.
Please use this as a guideline for your performance modeling efforts.
If you have any questions or concerns, please don't hesitate to let
me know.
Khaled Amer
President, AmerNet
Inc.
Architecture Analysis and Performance Modeling
Specialists
Address: 13711 Solitaire Way,
Irvine, CA 92620
Phone:
(949)552-1114
Fax:
(949)552-1116
e-mail: khaledamer@xxxxxxx
Web: www.performancemodeling.com
|