The New Cisco Nexus 2232PP Is Nice!

It Could Fix the "Build it, and They Will Come" Problem We Are Experiencing

Remember the catch-phrase in Field of Dreams - "Build it, and they will come."? The pretty much sums up our experience with 10GbE host connectivity over the past year in our new cutting-edge, 10GbE data center. We built the environment using Nexus 7000s and Nexus 5000s in the top of rack so we could provide 10GbE right to the hosts (servers, NAS, etc). But, we also incorporated Nexus 2000 FEX modules in each section of the data center for 1GbE connectivity. Even with some virtualization, 1GbE is still a lot of bandwidth; usually enough for most rack-and-stack servers. The 10GbE ports, even though there were more of them with the Nexus 5000s, were reserved for servers that really needed it (heavily virtualized rack servers, data base servers, blade centers, etc). The 1GbE ports were there for everything else. But something happened along the way; call it the unintended consequences of our good intentions. With 10GbE available to all hosts in the DC, suddenly, everyone "needed" 10GbE ports. Projects would come in for architectural design and the solutions architects would quote 10GbE servers. A good bunch of these servers didn't need it, but they still got 10GbE. Plus, our normal hosting hardware purchases for base virtualization inventory all started coming with 10GbE. Thus, we had a lot of demand for 10GbE ports. Our only recourse was more Nexus 5020s inside the same POD in the DC. That's pricey, particularly, as I mentioned, when most of these servers don't push 10Gbps, let alone 1 Gbps. So, I was excited to see the new Cisco Nexus 2232PP FEX module which provides 32 10GbE Ethernet ports in a single chassis (FCoE included too!).

I won't give out the pricing comparison between a N2232PP and a full Nexus 5000, but let's just say it's much cheaper. Now we can provide 10GbE ports at a much more cost effective level.

This simplifies our host ordering standards and makes people feel better because they too get a 10GbE port for their server. ;-) Of course, Nexus 2232PP FEXs aren't for every server. Blade centers and data base servers will still get 10GbE from the Nexus 5000s. Plus, since all traffic still has to traverse the Nexus 5000 (even for two hosts connected to the same N2232PP) is would not be right for applications that talk between each other a lot (however, I find these to be rare in enterprise data centers; most applications communicate with storage or databases, not each other). But, for commodity servers that are coming, by default now with 10GbE ports (the "build it and they will come" problem), the Nexus 2232PP is a great addition to the Nexus line.

One closing note. This is another good example of how competition is making our data center networking better all the time. Yes, maybe the Nexus 5000s aren't at the top of the performance charts right now, but there are other ways to compete. Adding this Nexus 2232PP functionality give the Nexus 5000 line a feature other vendors don't have right now. Good for Cisco!

More >From the Field blog entries:

Layer-0 Page for Network Design Templates

Nexus 5000 Only Supports 16 EtherChannels

A Graphical View of My Idea of WAN Acceleration and More Bandwidth

I Have Been Disillusioned by WAN Acceleration

WAN Carriers Need Better Bandwidth Pricing Models for Enterprises

I Got an iPad

  Go to Cisco Subnet for more Cisco news, blogs, discussion forums, security alerts, book giveaways, and more.

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Related:
Now read: Getting grounded in IoT