Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [802.3BA] MMF Extended Reach conference next conf call on September 4th

Hello Paul,

I've reviewed your presentation and have the following observations.

You have used the original version of hays_01_0407 for your base data. You
should be using the revised version which was updated in July 2007 with the
latest IDC forecast for X.86 server shipments - this reduced annual volume
growth from 10% to 6%. The amended figures are contained in the HSE
tutorial presented in Nov 2007 (copy of the amended figures attached).

Hays_01_0407 does not address aggregation of 10G server traffic but I'm not
sure about some of your assumptions in slide 4. Hays_01_0407 was based on
100G aggregation for 40G servers - a ratio of 2.5:1. You assume 10:1. This
would change the content of your slide 5. I would expect to see 40G
aggregation of 10G server traffic - a ratio of 4:1 - and I expect this will
use 10G LAG in the early years for economic reasons. We need to consult the
server system experts on aggregation levels and model this carefully.

If we assume that 40G will be used to aggregate 10G server traffic, then
taking 2015 as an example:

11 million 10G servers shipped (revised forecast) would give 1,833K 40 G
aggregation links using a 12:1 ratio and 2 links per server.

0.7 million 40G servers shipped (revised forecast) would give 117K 100G
aggregation links using a 12:1 ratio and 2 links per server (the same
figure as hays_01_0407).

Taking your 11% figure for backbone links in excess of 100m (as a %age of
total backbone links) we get 201K 40G links and 13K 100G links, or a total
of 214 XR links. If you extend this analysis beyond 2015 you get some
interesting trends. I've plotted this graphically in the attachment.

Best regards,   Alan

Attachment: Data Centre Link volumes over 100m.pdf
Description: Binary data