Blog Post

Why Network Virtualization Is Important

Stay on Top of Enterprise Technology Trends

Get updates impacting your industry from our GigaOm Research Community
Join the Community!

As part of its ongoing (and seemingly endless) battle with Cisco Systems (s CSCO), Juniper Networks (s JNPR) has announced a new box — the TX Matrix Plus, which makes a multichassis router out of 16 Juniper T1600 routers. Juniper first rolled out the T1600 in June of 2007; each T1600 had the ability to handle 1.6 terabits per second of data (800 incoming/800 outgoing.) So 16 of those boxes translates into 25 terabits per second of total capacity, or roughly 1,000 10 Gbps ports. This is a way for Juniper to catch up with Cisco’s CRS-1 router, which has snagged about 20 customers so far. The TX Matrix Plus isn’t due to hit the market until the end of the third quarter, but that doesn’t mask the fact that the demand for such routers is only going to increase as we put more and more information online.

Whether it comes via Google (s GOOG) or some social network startup, online data is exploding, and it’s putting new kinds of stress on our infrastructure. The economic mess we’re in means the infrastructure gets stressed even more, as companies and consumers alike turn to the web in an effort to squeeze additional efficiencies out of the system and save a few pennies. Such infrastructure is already under pressure as the demand for online video explodes — another phenomenon that that shows no sign of letting up anytime soon. (Related post: Why We Need Fat Pipes: The Top 5 Bandwidth-hungry Apps.)

Thanks to companies like Infinera (s INFR), however, it’s become easier for fiber network owners to provide more bandwidth in the networks’ core. And while Cisco and Juniper are providing routers to match those network speeds and capacities, the bigger, fatter networks also mean higher operational expenses for carriers. Which is where Juniper’s mantra of “Virtualization in the Network Core” comes into play.

With the TX Matrix Plus and JCS 1200, Juniper is addressing these challenges by enabling the hardware virtualization of highly scalable, adaptable core routers. Juniper core routers can be partitioned — on a per slot basis — into multiple virtual routers, each of which might represent services or network element types, and which can share resources such as interconnecting links and uplinks.

The logic is the same as that of the data center: Virtualization allows you to get more from the servers, at a lower costs. Ditto for routers. With Juniper’s TX Matrix (and related products) carriers can manage numerous large boxes as one single entity. Applying policies for QoS and other traffic engineering features makes sense when you’re talking about groups of routers, and doing it through a single, virtual entity makes it easier. Furthermore, network service providers can essentially use Juniper’s technology to create virtual networks. Light Reading explains it well:

Juniper is pitching its new core router as a vehicle for networking-as-a-service. A TX Matrix Plus setup would sit on the network as a pool of bandwidth that multiple parties could tap, letting a carrier lease out part of its backbone, in a sense. For instance, large carriers could make peace with over-the-top video services by giving them their own content delivery networks, which the carrier could even manage. The carrier could become the wholesale provider of a small operator’s network backbone, or lease out a network arm to a service provider that lacks a presence in a certain geographic area.

7 Responses to “Why Network Virtualization Is Important”

  1. Juniper’s new router might have some near term issues in the cloud world. The 100G solution uses 2 50GE processors – works in principle, but there is a packet re-ordering problem they’ll have to solve. And, it’s questionable if the 100G solution will be standards compliant – and if you have a non-standards compliant solution, are you locked in to Juniper? 100G is going to be necessary in the cloud world. Vendors have to get this solved – and soon to make virtualization a working reality.

  2. My point was that these ‘new’ ways of subdividing the ports and sub networks don’t materially add to the existing network management features. If they haven’t upgraded transceivers, there is no great urgency to have virtual ports. Gotta have one to take advantage of the other, unless of course, the facility is way under provisioned.

  3. I don’t think that the point here is additional link capacity or higher throughput. Virtualizing multiple high density routers is all about lowering operational costs and building efficiencies into future deployments. No one wants to configure thousands of ports individually – they want to configure a virtual network or pathway and have policy (QoS, traffic shaping, etc) apply to all of the ports along the path. Just like server virtualization technologies allow systems administrators to manage lots of physical server resources, network virtualization should apply to lots of physical router resources. As a concept this is nothing terribly new, but with the word “virtualization” stuck onto this concept it might get JNPR some good press :)

  4. Alan, you raise a good point, but the JNPR and CSCO boxes are going into core MPLS nodes. So it’s not really an issue of DWDM capacity, because the locations where they’re being installed already have it. But your 20% number is right – little need right now for a multi-chassis router in a typical 10,000 line central office.

  5. These innovations will never have an impact on link capacity, but may increase the opportunities for granular segmentation. At the metropolitan fiber level, fewer than 20% of the carriers have upgraded inter facility with DWDM transceiver cards. I think that before carriers spend money on fancy virtualized routers, that the link level upgrades have to go in. Most carrier network management apps already have underused virtual circuit features.