Implementing data center server virtualization and storage virtualization is no longer solely the job of the systems, data center or storage teams. Network managers are increasingly charged with the task of
networking the data center and storage systems, providing virtual machine management and troubleshooting virtual machines (VMs).
But there are challenges involved. Virtualization management strategies and tools are new, and networking pros are still learning what works. What's more, many network administrators don't have the training and preparation to work in virtual environments. Then, network managers must also find ways to break down IT silos within the enterprise in order to effectively manage and troubleshoot in virtual environments.
In an interview with Burton Group senior analyst David Passmore, SearchNetworking.com associate editor Tessa Parmenter asks the questions networking pros need answered about managing and troubleshooting networks in a virtual environment. To get a glimpse of Passmore's session at Catalyst, "Addressing Complexity: Networking for Dynamic Virtualized Servers and Storage," Parmenter asks why IT silos need to be broken down, what vendors are doing to help enterprises and how the challenges of virtualization present network administrators with an opportunity that could either make or break their careers.
Can you explain the challenges that network administrators face as they attempt to ready the network and management tools for virtual machines that work in accordance with physical machines?
David Passmore: Virtualization really affects the data center more than in other places. For example, if you look at a data center environment that's running VMware, VMware now allows virtual machines to move from one physical server to another, and somehow the network has to be able to track that. The network has to know that the virtual Ethernet address of a virtual machine has moved from one server to the next -- the same with the IP address of the virtual machine or of a particular application.
This is requiring tighter coordination between the networking people in a data center and the people who administer the servers, or the storage. Enterprises have to worry not only about coordination but about changing the workflows. Whenever a new application is provisioned within a data center, what sequence of activities is required now to make sure that the network connectivity is in place before the application boots up for the first time? These challenges are more significant now because data centers are very dynamic and virtualized. It means the networks have to be equally flexible.
That's right. I've heard that in order to make virtualization work, IT silos must be broken down – especially among networking, systems, storage and security. Is this unification actually happening?
Passmore: It's actually starting to happen because enterprises are running into these issues. Having totally siloed IT departments doesn't work anymore -- especially in the data center environment. The different groups are realizing that they require tighter coordination as they move to a more dynamic virtualized server and storage environment.
How long do you think it will take for IT silos to be unified?
Passmore: It's not like you flip a switch. A lot of it is people over time gaining knowledge and experience in areas outside of their comfort zone. Most people operating in one of these data center environments will tend to have a background of networking, or storage, or server operations, or security -- and then over time you'd expect that they'll start picking up knowledge in these other areas so that they can get a better big picture of what's going on and a better sense of what they need to deploy to create this next-generation virtualized data center environment.
Server and storage virtualization in the network
OS and virtual applications, who needs VMs?
Organize your networking group to stay above stressful times
Interop: Reinvent data center networks by smashing silos
Computer networking trends 2009 from David Passmore
Catalyst: New protocol for WLAN performance management and security standards
Are network administrators prepared to handle troubleshooting and virtual machine management?
Passmore: No. They're not prepared. In fact, many of them are still getting off the learning curve -- more or less still trying to learn how to spell "virtualization."
More seriously, they're having to become familiar with new technology -- like virtual machine hypervisors -- and learning that they may now find themselves saddled with virtual Ethernet switches that are implemented in server software -- something they never had to deal with before. There used to be physical hardware Ethernet switches; no one ever thought of an Ethernet switch as being something that you could implement in server software.
What specifically are IT networking professionals missing, and where do they find training in virtualization?
Passmore: They're having to learn about these things from the same places as their colleagues in storage management or server management.… [Network engineers will have to] become familiar with a different set of vendors than they're used to. The networking people in the data center have to know quite a bit about server vendors and storage virtualization vendors. It's a whole new set of players that they have to become familiar with.
What is your assessment of the efforts network equipment vendors are putting into their products to help network managers deal with dynamic data centers? Are technologies like Cisco's Nexus 1000V virtual switch moving us in the right direction?
Passmore: Yes, in fact this is great news for the network equipment vendors, because for the last decade they've produced Ethernet switches for large enterprises, and quite frankly, these were relatively mature products. A lot of enterprises have 10-year-old Ethernet switches that they've been perfectly happy with because they do the job. But now, with these dynamic virtualized data center environments, it requires a new class of products. In particular, Ethernet switches that can, for example, track the migration of virtual machines. And rather than viewing that as a challenge, it's actually a huge opportunity for the network equipment vendors because now they have a whole new set of products to sell the enterprise. So they're pretty excited about this. It kind of breathes life into what was otherwise a pretty mature market space.
How long do you think it will be until there are solid virtualization product choices in the market?
Passmore: Right now, the networking products that can support virtualized data centers are relatively immature. Some of the standards that are necessary to make this happen, like converged enhanced Ethernet, probably won't get finalized until at least late next year. So, right now, it's really time for enterprises to be trialing some of this new equipment and gaining some more experience with it, without necessarily whipping out their checkbooks and putting in place a huge new production network. So, in other words, it's still early days.
Depending upon the kind of enterprise customer you are, certain vertical industries are more risk averse than others. Some of the more competitive industries may be forced to migrate to this relatively quickly because they have significant cost or capacity pressures; whereas others, like parts of the financial services industry, may want to lag behind because they're far more concerned about the security and stability of their infrastructure, and they'd rather wait for other people to be pioneers -- like the engineering and scientific computing verticals.
What does it mean to say that a dynamic, virtual environment reflects the "risk appetite" of an organization?
Passmore: Simply the fact that for enterprises to install virtual machine hypervisors so that they can, for example, run multiple virtual machines inside a single physical server, does entail more risk because you're worried about the security of multiple servers sharing the same hardware -- potentially being able to access each other's data. It means you need to be more careful about the more complex configuration management that's required. So it's definitely riskier than giving each collection of applications its own dedicated server.
On the other hand, the rewards are also greater in that you'll be able to buy a third as many servers. You may have the flexibility to load balance so that you can better accommodate peak demands for application processing. It allows you to get by with less storage than would be required otherwise. A lot of that is being driven, pure and simple, by economics. Going with virtualization technology is often a way to save money.