How would you assess the current state of challenges virtualization poses to networking? And what sort of progress have vendors made in helping networking pros deal with these challenges?
Also, one of the things virtualization drives is the shift from direct attached storage to network storage of some type, whether FCoE or iSCSI. To take full advantage of virtualization requires network storage, so for those who haven't made that shift, it will cause them to implement some form of it. We see a lot of folks rolling out directly to iSCSI, and we're also seeing some renewed looks at NAS -- because of some of the changes in the way virtualization works now, you can take advantage of NAS and SAN.
In terms of how vendors have helped network pros, a series of protocols for desktop virtualization have been released by vendors. VMware has a PC over IP, which allows better graphic applications to be run in a VDI environment; Citrix has its new "high definition experience" (HDX) protocol, based on independent computing architecture (ICA) but adds other capabilities. So from a network perspective, you have to understand the protocol, and Microsoft has just announced its new Remote FX, a new graphics protocol based on the Calista acquisition. And the latest hardware has been made with network admins in mind. The newest release of vSphere allows virtual switches that are the same software as a physical Cisco switch, making life easier on the Cisco team.
And then there are the virtual switches. What happens when I consolidate from a physical to a virtual workload is multiple virtual workloads. So if server A used to talk to server B, and it went through a physical switch, I had to do all management through a physical switch; but if they're on the same server, now I have to go through a virtual switch and consider the management and security issues with the virtual switch.What kind of application delivery challenges can result from using server virtualization?
There can definitely be performance issues. As users move from the first wave of server virtualization (where people had maybe 20% virtualized) and move on to the next wave, they start to virtualize more mission-critical applications that have higher performance requirements. As a result, there is a higher demand for the proper troubleshooting tools to get performance back to an acceptable level. So it drives the need for tools that address the network's role but also tools that address an application that is running virtualized inside a server across the stack. There's a whole new set of tools that address these issues in a virtual infrastructure. How does desktop virtualization fit in?
Users are making the jump to desktop virtualization -- they've done server and are moving to desktop. It has some dramatic impacts on the network because it's shifting the location of the desktop work from the PC in front of the user to back in the data center. The communications now go all over the network, so that can have a significant impact on network traffic, and depending on the implementation, there are different impacts on the protocol layer.
In terms of protocols, VMware has a PC over IP, which allows better graphic applications to be run in a VDI environment; Citrix has its new HDX, and Microsoft its new Remote FX. All three of these protocols from the major vendors improve the user experience for desktop virtualization.
But there are other implications with desktop virtualization. What does it do to the network? What do these protocols consume? LAN or WAN? HDX can run over a LAN with a good user experience, and remote FX probably will not by itself, but Citrix and Microsoft are doing some work to allow combinations for these capabilities. PC over IP requires different bandwidth than the others, and different users need to look at the different aspects, including WAN optimization.Can you discuss managing dynamic virtual machines on the network and avoiding VM sprawl?
One of the big benefits for advanced virtualization is that it allows things to be much more dynamic. Separating the virtual from the physical allows movement of workloads much more easily within the data center environment, so as I'm trying to better manage my data center, I want to move a workload from one physical server to another for balancing, to patch my servers, and so on. So the data center becomes much more fluid and dynamic in the management of those workloads, and because those workloads require connectivity to other places, it requires a more dynamic nature of the networks as well. As I move the workloads, I need to move the networks and the connectivity that goes with it, which might require automation. Do you have any advice for breaking down the silos within an IT organization to get the network, storage and server people to work together?
I think virtualization is another one of those technologies that is forcing a need for better communications, because as we're virtualizing, we're virtualizing everything, and all of the pieces need to talk to one another -- networks, storage, server, desktop. As everything becomes geared toward private clouds and eventually the public cloud, it will require all of the groups to work more closely together, so we encourage cross-functional teams to be working together. Get the teams to talk so the server guys don't just go off and implement server virtualization and cause problems for the network guy.
And then there's another aspect of the virtual switches. The newest release of vSphere allows virtual switches that run the same software as a physical Cisco switch. So before, when I managed my virtual switches, the management switched from the management team to the VMware team. Now, because those switches are like the Cisco switches, it can switch back to the Cisco team because some switches are physical and virtual. This causes some organizational changes as well as enabling some better management from a technical standpoint.