Reflecting on last week’s Interop in Las Vegas I found myself thinking as much about what wasn’t there as what was.  While I believe the official reports that claim attendance gains over last year, the show just felt small.  More importantly, it felt uninspired.  Here we are in the midst of a major transformation in the way we build computing systems, led almost entirely by VMware, Amazon, Microsoft, Google – and without major influence from the networking industry. 

The network is behind, way behind, when it comes to delivering the strategic benefits of cloud computing in terms of dynamic, flexible movement of workloads among computing centers.  And from what I saw at Interop, its about to get worse – potentially, much worse.  The hypervisor vendors and associated tool providers have done an amazing job of automating the process of instantiating and moving workloads – as long as they all stay within a common VLAN.  That’s a potentially big limitation even within a single data center, and it’s a show-stopper for moving virtual machines between data centers and, dare we dream, between private and public clouds.  

Delivering flexible VMotion beyond the confines of a VLAN requires a significant amount of tinkering with the network – IP addresses, DNS records, ACLs, firewall and load balancer rules – all of these need to be “moved”, just like the virtual machines.  So who’s building the tools that can move virtual networks as elegantly as Virtual Center can move VMs?  I can tell you this, if what I saw at Interop truly reflects where they are, then its pretty clear that the major networking vendors aren’t doing it.

I participated in a panel session at Interop titled “Why Networking Must Fundamentally Change”.  The intent was to discuss how networking must evolve to deal with the new challenges of virtualization and cloud computing. The first part of the session was a discussion among senior representatives from Cisco, HP, Dell, Juniper, LSI Logic, and Arista, there to present their visions of how to build networks for the next generation data center.  This was followed by a discussion among network automation vendors, including Egenera, Linesider, and Infoblox (my company).  I don’t purport to fully understand what the networking folks were saying, but I do know that what I heard scared me.

Several of the vendors are talking about flattening the data center network, i.e. turning it all into one big switch so that packets don’t need to traverse layer three on their way from blade to blade.  From a sheer performance and latency perspective that makes perfect sense.  But from a flexibility perspective, it could be disastrous.  From what I could tell, there’s little if any agreement on exactly how to extend these layer 2 fabrics between data centers.  Well, that’s not exactly correct:  All of the vendors (with the sole exception of Arista) seem to agree that it’s perfectly fine for each vendor to do it their own way.

Yes, that’s right:  At current course and speed, you’ll be able to get all of the cloud bursting and DR and resilience that you want, as long as you buy everything from one networking vendor and/or use cloud providers that also use the same networking vendor as you do.  It was other-wordly to be hearing this at, of all places, Interop.  There should have been a riot, but there wasn’t.  In fact, during the 1-1/2 hour discussion, not ONE of the networking vendors even uttered the word “cloud” – and no one seemed to care.  (Well, I did).

Look, I’m as skeptical as anyone about the wide-eyed claims being made about the coming cloud utopia.  But it seems a real shame that the networking industry has barely begun to tackle the issue of developing new standards and tools that enable real network automation that will work across vendor infrastructures.

Why would networking vendors want to do anything other than push proprietary architectures that effectively lock in their customers?  I assert that if being too open is bad for business, being too closed could be worse.  Without open standards, starting with IP, we might still be slogging away with IPX and X.25 and DECNET and Appletalk and – gulp – SNA.  Miraculously, even though SNA and the other proprietary networking standards died away, companies like Cisco and F5 and Juniper (and Google and Yahoo and Salesforce.com and on and on) were created, IBM grew substantially, and we’ve seen unprecedented gains in productivity and social changes that have improved lives and changed governments.  All thanks to a little networking standard.

There’s another big issue here:  Unless you’re about to dump your existing data centers and build them all new, you’re going to have to automate what you have now, which in many cases today is an L2/L3 monster still managed with spreadsheets and custom scripts.  If you don’t automate, your network will become the chief constraint on your ability to take advantage of virtualization.  

Therein lays a great irony:  Networks need to be automated now.  And if the problems of moving virtual networks can be solved within today’s data centers using existing L2/L3 networks and all of their associated security and load balancing and related systems, then it’ll be much more likely that those same solutions will work across data centers.  And that’s a path to realizing the real benefits of cloud computing.   

By the way – there *are* some standards being developed that go a long way towards automating the network.  Two that come to mind are new standards such as LISP, which decouples identifiers from location and makes networks more portable (https://www.cisco.com/web/about/ac123/ac147/archived_issues/ipj_11-1/111_lisp.html), and IF-MAP, a standard from TCG which enables dynamic, real-time coordination and configuration management among heterogeneous products from different vendors (http://www.infoblox.com/solutions/overview-if-map.cfm).  End users need to learn about these technologies and their implications, and ask vendors if they plan to implement them - and avoid a new era of lock-in that could make us nostalgic for the days of the mainframe.