News Stay informed about the latest enterprise technology news and product updates.

Interop preview: Data center network design a hot topic

Data center network design and new network management strategies will be top of mind at Interop Las Vegas, according to networking track chair Jim Metzler of Ashton, Metzler & Associates. In this Q&A, we ask Metzler to give our readers a preview of what to expect at this year's show.

First off, what do you think will be the hot topic among enterprise network vendors at this year's Interop Las Vegas?

Jim Metzler
Jim Metzler
I have a two hour and 15 minute session entitled "Why Networking Must Fundamentally Change." I think it was two years ago we had a session entitled "Is There Anything Left to Say about the Local Area Network?" Now, we're talking about why networking must fundamentally change. This is focused primarily on the data center network [design].

In the old days, like last year or the year before, we all knew data center LAN design: Access-distribution-core: Layer 2 here, layer 3 there. It's been well understood for the last 10 years. Now, everything is in play and there's a lot of discussion of collapsing that into fewer layers. Of course, you could argue that before we collapse it, we've added another layer with the virtual switch inside the virtualized server, so we ended up with a four- or five-tier architecture, depending on how you want to count.

Look at some of the documentation coming out of VMware. If you want to move a VM [virtual machine] around with vMotion, it has to be on the same layer 2 VLAN, and it can't take more than five milliseconds. That's not long. So there's been a lot of discussion about how we need to flatten data center LAN and go back to a layer 2 domain. Well, vendors don't agree on that. There's a convergence going on of Fibre Channel and Ethernet. So you're seeing a convergence of technologies with Fibre Channel and Ethernet and switches and servers.

So how are these technologies converging? What's happening with the basic structure of the data center LAN? And do we really need these guidelines from VMware about everything on the same VLAN and only 5-millisecond round-trip delay? If that is the case, then you can't really move VMs very far.

More data center network design resource and Interop Las Vegas previews
Power and cooling consideration in data center network design.

Can you take FCoE too far in data center network design?

p> How important is the race for highest density 10 GbE core switches to data center network design?

Interop Las Vegas preview: The future of application delivery appliances.

Computing has become virtual and dynamic, and everything else is struggling to keep up with it. So the discussion that should be coming out of Interop and continuing into next year is: "OK, how is the network going to change?" And go figure -- Cisco, Juniper, HP and the rest won't agree on that. So what do you want attendees to glean from this deep dive into data center network design?
One of my goals with that session is to identify where there is agreement and disagreement among the major vendors so that people in attendance can say, "OK, I feel safe about this." But they can also identify … the areas that they need to watch more closely. I think most major vendors will agree on a general flattening [of the LAN]. I think most major vendors see long-term Fibre Channel and Ethernet convergence.

Then some people think that you're going tie a VM to a port. Others don't think so. HP has switching on a NIC card as an option. Once you [understand at a] high level that we need a flatter network, I think the particulars of what that looks like are going to be very different. You have an Interop Las Vegas session on breakthrough network technologies. What's the goal of that session?
When you think about the standard session, it's an hour long. Well, there is stuff going on that's really interesting but not enough to take up an entire hour. Take Talari Networks, for instance. They say MPLS is really expensive -- $1,000 per megabit per month in the U.S. And if you want to go to 100 branch offices, that's pretty pricey. You can get Internet access but most people don't feel comfortable doing that with inexpensive DSL, so they get T1 access or something like that, and that gets it down to $500 or $600 per month. If they feel comfortable with DSL or cable, you're talking $15 to $20 per month, but it's not quite as reliable. But what if you put two of them together? DSL and cable -- there's not too much chance of both of them being down. And you're spending $50 or $75 per month. What if you can load balance those two and get really good performance? I see that as potential breakthrough technology. Enterprises are not necessarily going to do that in all of their locations, but they may want to do it in some of their locations and bring costs down. It's a fascinating topic. It's worth 15 or 20 minutes but not necessarily an entire hour.

I also have Martin Casado from Nicira. Nicira is a startup LAN switch vendor. We haven't had a lot of those for a while. For the longest time, the LAN switching market was very staid. Well, I've got startup Arista on the panel for "Why Networks Must Fundamentally Change," and I've got Nicira on this one, and they have some ideas about the LAN switch that are just a little bit different from Cisco's and Juniper's.

It gives me a chance to expose ideas in quick hits. I'm not saying that if people walk in, all five will resonate with them. But chances are, two or three will. It's a good chance to get exposed to ideas bubbling on the sidelines. What can attendees expect to learn in your session on Advances in Network Management?
Same thing as the breakthrough in network technologies. I want to give people a chance to talk about a technology or an approach to management that normally wouldn't fill a total hour. Look at ExtraHop, one of the vendors there. ExtraHop's take is that the technologies have really changed over the last eight to 12 years in terms of Moore's Law and what you can do. So I can do deep packet inspection of all packets and get much better management information than with some other techniques. Here's a startup out of Seattle that is going to apply that to network management.

You have NetScout, EMC and CA. We've always had this push-me, pull-me in IT of best-of-breed versus fully integrated solutions. If I'm ExtraHop, I'm going to come out and talk about my best-of-breed solution. If I'm CA, I'm a multi-billion-dollar company that's been acquiring lots of companies over the last five or six years… same with EMC. I'm going to talk about fully integrated solutions. NetScout is closer to ExtraHop along those lines, focused on best-of-breed versus manager-of-managers-type things.

Dig Deeper on Data Center Network Infrastructure

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.