•Optimizing R&D effort is definitely a good thing – Did we learn anything from 10GE? • Should we have a transceiver MSA before the standard is complete? • Are the components cost-effective and manufacturable? – 40GE doesn’t need to double R&D Effort but new chips...
Must not delay 100G
If we continue talking about 40GE, it will delay 100 GE Agreeing to add 40GE instead of endlessly debating will lead to faster standards completion. The better question is will 40GE delay the standard significantly or even measurably. Sun showed that there is minimal effort...
What are the number and type of PMDs required?
• Same as 100GE – Must skip the 10GE PMDs designed for OM1 Fiber (LRM) Another question is what transceivers are being standardized? • For multimode applications, the QSFP should work fine, is complete and is in preproduction at 20Gig (4 X 5G / lane)...
Show more end user support
• United Airlines and IBM •Brocade is an end user and we would like to use 40GE links between switches and to LANs to create File Area Networks (FAN) • Most 40 GE customers are more cost conscience than 100GE users and can’t afford to...
What is the relative cost between 4 x 10 G and 40G?
• Not the same if 40Gbase-SR4 is defined to go 100 meters while 10GBase-SR must traverse 300 m – Estimates are 10-50% less cost at 100 meters Cost isn’t everything. Users will buy 40GE over 4x10GE because of: • Ease of Management – it’s one...
Where would 40GE fit in 2010?
40GE would compete with 10GE and could increase sales of 100GE 40GE could give better value than 10GE with only 33% more BWLP but 4 times the bandwidth • Majority of new backbone links would be over new fiber in the core of the network,...
Why is 10GBase-SR difficult to make?
10GBase-SR put great constraints on transceiver designs • 300 meters * 10G = 3,000 MHz*km bandwidth length products (BWLP) of link – 1GE only has a bandwidth length product of ~1,000 MHz*km – OM3 fiber has a BWLP of 2,000 MHz*km, so the BWLP of...
Why is 10GE so slow (at being adopted)?
Extract from Finisar proposal to 8GFC Fibre Channel (http://http://www.sfpcables.com/) : Practical 10G VCSELs to date have been small aperture MM devices. – Yield VERY poorly even to largest allowed spectral width – This is largest single cost driver for 10GBASE-SR Practical TX design almost impossible....
10GE Can’t Compete with 1GE
According to Dell’Oro, the average selling price (ASP) of a 10GE port was 31 times higher than a port of 1GE in 2006 – Price ratio between 1GE and Fast Ethernet was 4.8 times 10GE is in a different league than 1GE 1GE ASP approaching...
Recap of 10GE Adoption
Dell’Oro reports and forecasts 10GE port shipments every six months. This graph shows forecasts and predictions from the three latest reports. Predicted 854k ports in Jan 06 Report Reported 318k ports in Jan 07 Report 526k Ports shipped by Jan 2007 More ports should ship...
