Let me get this stright...3 of your brightests engineers leave your company, you give them $100M to develop a product..a year later you buy these guys out for $750M!...this is American dreamed exponential...I need to talk to my boss about spinning me out and in!...I will take only 10% of what Insieme folks took ;-)
I have not designed a networking ASIC for at least 10 years but still remember few things on the topic ;-).You probably need 1 Tb/s aggregate bandwidth (or more) so with 10 Gb/s per pin (differential) that requires 400 pins. Plus ground, power, control etc. Can be 1000+ IO pins at the end. Plus billion little muxes inside the core. Requires careful engineering, probably cost few million $ to tape-out but no smart packet processing as it is done today (famers, packet processors, etc) which will be done elsewhere.
The bottom line is that it's still a matter of N-to-N connections and no software can solve that problem. BTW, in today's complex routers there is lots of software already so the software component is not new, it is just executed somewhere else so networking can be controlled by the box operator not by the box manufacturer.
At the end all the software can do to help get packets where they need to go is simplify the core switching problem to something like a cross-bar, you can't go simpler than that! BTW, all switching concept are very elegantly explained by Carl McCrosky is our Wiley book "Network Infrastructure and Architecture: Designing High-Availability Networks".
Looking forward to your interview with Insieme folks! Kris
SDN starts from openness. The primary goal is to ease management and potentially improve network utilization. To me, a consortium shall be formed and a set of standards shall be agreed among different vendors to ensure active routing/switching reconfiguration can be in a larger scale. I can understand the benefit of proprietary solution to a company. I believe the power of standard and agreement to a bigger scale.
A Book For All Reasons Bernard Cole1 Comment Robert Oshana's recent book "Software Engineering for Embedded Systems (Newnes/Elsevier)," written and edited with Mark Kraeling, is a 'book for all reasons.' At almost 1,200 pages, it ...