x
Comms chips

AlcaLu: Going With the Flow

CHICAGO -- Light Reading's Big Telecom Event -- Alcatel-Lucent's Basil Alwan spent part of his keynote this week grappling with what networking functions and elements will get swept away by the tide of virtualization and which can't simply be replaced by software and cheap servers.

Alwan, president of IP routing and transport for Alcatel-Lucent (NYSE: ALU), stated that there is a "tremendous pressure" of change on vendors and carriers as SDN and NFV shake up what networking is and can be, but there's also a "tremendous opportunity."

Some things, however, will not change. "Software defined or not software defined, we still need big routers," Alwan told the audience.

Surfin' BTE
Alwan takes the stage.
Alwan takes the stage.

Lest you think that the vendor guy is simply a King Canute trying to hold back the software tide and shore up his margins, Alwan gave examples of where the company has realized it has to work with third-party platforms and software.

For instance, AlcaLu's SDN spin-in, Nuage Networks , has built its platform on hardware from Arista Networks Inc. and HP Inc. (NYSE: HPQ), among others. The company's Cloudband virtualization platform pulls in third-party elements, too. (See How Alcatel-Lucent Set a Telco Cloud Example and Alcatel-Lucent Expands CloudBand.)

The key, Alwan says, is knowing which functions of the network can be abstracted with software and x86 servers and which will still require dedicated network elements. "I like to call it network function optimization, rather than virtualization."

Specifically, he thinks that elements like the mobility management entity (MME) in the packet core could be substituted with software on an x86 platform. Hardcore IP routing elements designed with SLA requirements in mind, he says, are built around dedicated network processor performance that PC-derived x86 silicon simply can't match yet.

"It'll be five years, easy," before performance starts to get good enough, Alwan said.

So, for Alwan and AlcaLu, this is what the idealized evolution of the network will look like.

Free Flow
And routing transport IP shalt live in harmony with SDN and NFV.
And routing transport IP shalt live in harmony with SDN and NFV.

What say you, readers?

— Dan Jones, Mobile Editor, Light Reading

dwx 6/18/2014 | 3:12:54 PM
I would agree I agree core routers aren't going anywhere anytime soon.  You can't find a 100GE interface on any "bare metal" network device even and certainly not anywhere near a server.   The only way to interface something like that into a server is put a NPU on the server and then you might as well be using a router which consumes less power.   By the time servers catch up we will be using 400G interfaces off of routers in high bandwidth applications.  

I saw a picture once of how many racks of servers you would need to process the amount of packets a full-tilt single Cisco NCS6K could (8Tbps full-duplex) and it spanned more than a couple rows.  Of course this is a marketing slide but it's not really all that far off.  

As you start moving NFV functionality into the cloud for functions where it makes sense like virtual CPE, FW, etc.  you need bigger pipes to connect everything together.  The inter and intra-datacenter traffic will continue to grow at speeds outpacing x86 certainly and even commodity hardware unless the commodity hardware folks start producing higher speed NPUs.   Juniper/Cisco/ALU have all had 100G+ NPUs for some time now. 
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE