Infrastructure design and construction will be a major topic of discussion at Interop NY this week, especially as networking, data centers and storage continue to evolve and new technologies are introduced. To get a glimpse of what Interop attendees can expect from the conference's track on infrastructure, SearchNetworking site editor Chuck Moozakis spoke with track Chair Eric Hanselman, chief analyst at 451 Research.
What are the hot topics at this year's New York Interop?
Eric Hanselman: Things have picked up a notch. There is nothing like John Chambers [Cisco Chairman and CEO] doing the lead-off keynote to get everyone a little more excited. There is a lot of anticipation around what's happening with Cisco, and all sorts of rumors about what will be debuted when. But simply having John's position on where Cisco is headed and his perspective overall is useful. When you are in an environment where you have such a dominant player in the market, [Cisco] has a very strong position in how they are influencing what we all are doing in networking. There is also a little more running time around the shifts that are happening in networking; the technologies [developed over the past several years] are getting out there, and now we have some running time on what's been theoretical networking. People are turning out to get a better understanding of next-gen technologies.
Convergence is always a topic that drives a lot of interest. What does convergence actually mean?
Hanselman: What we have is convergence happening on a number of different fronts. We have platform convergence happening, in that we are getting integration of systems for deployment that are pulling compute, networking and storage together in a single platform. We have technological convergence on the networking side. We have storage and data networking converging and that is probably one of the more active areas in terms of network integration. We've seen virtualization starting to drive both of these to new levels as users begin to sort out how to get reasonable levels of efficiency to cope with the higher levels of utilization we are working with. Most of what we will be focusing on at the infrastructure track is taking a look at what's happening with storage networking convergence, and what a lot of those integrations really mean. It's kind of a scary topic for many folks. These were realms that were relatively distinct and discrete. But when you look at starting to merge those traffic types together, it makes networking folks concerned about the volume of traffic that storage interaction generates and it makes storage teams concerned about performance requirements that they need to get out of a converged network.
There's a session on application delivery in the new data center. What will that cover?
Hanselman: We will be taking a look at application delivery controllers and looking at a fairly dramatic shift in capabilities. We've thought about ADCs as being the next fancier name for load balancers. What's happened is that we are looking at much more complex application environments. The ADC is now the place where you are starting to do all sorts of things: identity management; Quality of Service on delivery; being able to look at service-level delivery capabilities; security; layering all sorts of things into what had been a fairly simply box in the days of yore -- just making sure there was some level of resilience in terms of performance. Now it's become this nexus where a whole set of different capabilities are brought to bear on an end user's interaction with the back end infrastructure that's serving up the data they are trying to get to.
The Wi-Fi session covers why the industry needs to embrace new Wi-Fi standards. Are organizations reluctant to deploy new technologies?
Hanselman: What is out there now has made significant changes [in the enterprise]. 802.11n has been out there and has been widely deployed, and we are now having a dramatic change in moving to 802.11ac. What is happening is that we are in the cycle where we made the move to n, and now we are getting the base level silicon for 802.11ac, so we are at a point at which it's starting to get to be a relatively small cost step to be able to go to this newer, faster and better technology. But underneath this we have some fundamental issues we've always had with Wi-Fi. One of the difficulties with Wi-Fi is that it's a place where you have to merge a lot of technologies together. You have the RF angle, you have the backend capacity angle and you have the endpoint capabilities, so really making Wi-Fi in whatever standard you happen to be using function well is a challenge. You also have something that -- traditionally in networking -- you've been able to focus on using a pretty narrow domain of skills to make it work right. Now, with the shifts to the next step, these are areas in which you really have to master some fairly complex arrangements to make it work well. Just because you have five bars of Wi-Fi doesn't necessarily mean you'll get a solid connection. Wi-Fi continues to be the highest priority in terms of budgets because organizations are still trying to figure out how to manage this huge swell of mobile devices. We're in that transition from wired to a mostly unwired environment.
Talk a little about the sessions on network function virtualization.
Hanselsman: If you look at network function virtualization, we have stuck a lot of functionality in the network, and through that we've crushed a lot of different things into what should have been a relatively simple box. As we start to virtualize network connectivity, now we can start to break apart this great mass of different things we've lashed together. These are areas like: Should we have really had a full OSI stack of functions all running in one physical box? With network function virtualization, we're starting to pull apart some of these pieces. Network function virtualization lets us get beyond the appliance model and look at all these capabilities in more atomic ways so that we can scale them or apply them in different ways -- even down to different flows of an individual connection. Now we can do useful things that break apart what had been a broad collapsed set of functions sitting in one great huge box. Once you break these apart, now you have a number of different ways to address these problems. That said, the sophistication of these capabilities that we are starting to bring isn't still up to speed with the individual system-based capabilities that we've been working with. One of the big questions is how to manage this all. The next big decision point or point of inflection around who controls which aspect of network is really looking at the service delivery and service chaining functions. Who is it that inspects the traffic, decides how it's going to be managed and gets it off to the appropriate pieces to do whatever that particular performance function is? There are a number of vendors circling around this space. Everyone is jockeying for position for where that control is headed next. We've gotten sophisticated about orchestration around compute and storage, but we are really trying to figure out how to integrate networking into that whole IT/orchestration piece. As we head through the next year, we are really seeing some dramatic changes in what the options are and really what can be brought to bear to tackle what has been the third leg of the IT service delivery piece.
Cisco switch alternatives: What will those sessions cover?
Hanselman: As we start to recompose how we build networks, we have a lot of different options. We are hoping to sort out what all these options are. There are a lot of new players in the field who are coming in with ways to look at how we handle all the things that networks have typically taken care of. We've had the traditional networking vendors for ages. Now some alternatives like OpenFlow and virtual networks have gotten to a point of maturity where you can do some meaningful things with them. There are a lot of different options. And they are interesting options in different ways. We are starting to see a lot of specialization. With the rise of merchant silicon, we now have a set of silicon building blocks that make the task of simply building a pretty fast switch really easy and that's allowed a slew of new companies to start to innovate on top of the merchant or custom silicon that they are building. Once we get into this much more generic platform, it allows companies to let their imaginations loose on how they can handle what are more complicated questions that we simply haven't had the infrastructure to be able to work with in the past.