On Nov 3, 2011, at 5:33 PM, Tim Preuss wrote:
Bruce,
I suppose you have seen this article or ones like it. If you have time,
I am interested in your reaction. Does this foreshadow a major change?
The article mentions OpenFlow also which has been discussed at the Internet2 Joint Techs meetings and also at InterOp this year.
I haven't really been impressed with OpenFlow yet, it reminds me too much of a company that was promoting an ATM based box that would set up flows but be managed by a box that was looking at the traffic as IP traffic rather than ATM traffic. It was a bit kludgey. Google helped me, it was Ipsilon.
http://ccr.sigcomm.org/online/files/p81-v36n3p-minshall.pdf
I do see why OpenFlow is interesting for people working in network research such as new protocols etc though.
It was interesting to see that NEC and HP and a startup had more support for OpenFlow than I would have expected.
The article also mentions Vyatta and I believe is correct in how it categorizes Vyatta as targeting the smaller devices at the edge of the network.
There have been interesting threads on discussion lists like NANOG about the power of a commodity box to move bits.
http://mailman.nanog.org/pipermail/nanog/2011-September/thread.html#39973
http://mailman.nanog.org/pipermail/nanog/2011-September/040547.html
http://mailman.nanog.org/pipermail/nanog/2011-September/040103.html
http://mailman.nanog.org/pipermail/nanog/2011-September/040013.html
http://mailman.nanog.org/pipermail/nanog/2011-September/040022.html
http://mailman.nanog.org/pipermail/nanog/2011-September/040513.html
http://mailman.nanog.org/pipermail/nanog/2011-September/040551.html
http://mailman.nanog.org/pipermail/nanog/2011-September/040555.html
We have several $80 Mikrotik RouterBoard 750Gs and I'm pretty impressed with the capabilities. It's sort of the model mentioned in the article with a software based router controlling a device that is also a hardware based switch. So you can switch between ports using the switching hardware or you can route between ports and there is just a general purpose CPU running software that is doing the routing function.
At the other end of the spectrum you have the Cisco ASR and Juniper's new chips.
http://www.cisco.com/en/US/prod/collateral/routers/ps9343/solution_overview_c22-448936.html
http://www.expresscomputeronline.com/20100315/expressintelligententerprise27.shtml
One way to look at it is that it's the age old question of specialized circuit (nowadays specialized chip) vs a general purpose CPU that is driven by Moore's law and economy of scale.
At one time bowling machine controllers and auto scorekeepers were driven by circuit boards with discrete logic chips. But as general purpose CPUs got cheaper and more powerful it made more sense to make the bowling machine controllers out of CPUS and software rather than custom hardware.
Another way to look at it is that in addition to servers being virtualized portions of the network are being virtualized also. If you have a VMare host running 12 guest machines it can be configured for the 12 guest machines to talk IP to each other over a virtual network multiple virtual networks on the VMware hosts. And the virtual network may have multiple vlans, ACLS and perform routing functions etc. All of those network functions are essentially running on commodity hardware.
I think that for now in the core of the Internet the specialized hardware routers are safe. There is even worry that they won't have fast enough affordable memory to handle the increase in the number of routes.
At the edge routers based on commodity Intel/AMD chips and running routing software are viable. In between the core of the Internet and the edge such as the core of NDSU's network there will be mix depending on people's needs. We still put high end switches in our wiring closets with dedicated hardware to switch ethernet packets. We use enough layer three features like IPv6 QoS and IPv6 ACLs for security, IGMP and MLD snooping etc that it is worth it for us. But in another environment and ethernet switch with commodity ethernet chips (vs custom Cisco chips and vs commodity general purpose CPUs such as in a PC) might meet all of their needs just fine.
At the same time that general purpose CPUs are getting more powerful Internet bandwidth demand is growing around 40% per year. Our backbone at NDSU is 10 gig and that is meeting our needs just fine but there are places in the Internet backbone and perhaps huge data centers that needed affordable 100 G ethernet last year.
So I think there will be changes, and there already have been some but the changes won't necessarily be everywhere in the network.
http://gigaom.com/cloud/how-an-extinct-zebra-could-upend-the-networking-market/
Thanks,
Tim
--
_________________
Tim Preuss
Information Technology Instructor
1900 28th Ave South
Moorhead, MN 56560
tim.preuss@minnesota.edu
www.minnesota.edu
218-299-6614
One College. Four Campuses. Online.
---
Bruce Curtis bruce.curtis@ndsu.edu
Certified NetAnalyst II 701-231-8527
North Dakota State University