The movement of virtual machines between servers is normally used for routine maintenance like upgrading hardware...
and load balancing. This normally occurs within a single data center. It becomes even more challenging to plan for disaster avoidance in the event of a local catastrophe, because the network manager has to develop an infrastructure for redirecting IP addresses across a WAN.
The main challenge in this scenario, according to Patrick Lin, senior director of product management at VMware, is that the network manager would have to stretch the VLAN across multiple subnets, which is not yet very common.
Moving virtual machines (VMs) between physical servers is extremely common. Anyone with a production environment will need to understand the impact of moving VMs and create workarounds for them. The major problem is that the IT infrastructure becomes more dynamic. Simon Crosby, CTO of the Virtualization and Management division at Citrix, noted that the challenge is configuring the network and storage services that are scattered around within the server itself and then within the components of the infrastructure such as the switches and firewalls.
"As VMs move across subnets, just the movements of the VMs can cause headaches around keeping track of where resources are and who has access to them," Stephen Elliot, enterprise systems management analyst with IDC, explained. "It is as much about change control as change management. The network manager has to make sure the network teams have visibility into the movement of applications and servers. Network managers have to develop more effective ways of tracking the location of applications than was required with physical servers."
An important protocol in this domain is the VLAN protocol or IEEE 802.1v, which allows the networking manager to create a virtual network across multiple switches. This can help reduce the complexity at the service layers for doing such things as moving an application from server A to server B. But it can become more complicated when trying to maintain port configurations during the migration.
Crosby said companies like Scalent Systems have developed solutions for helping to automate the configuration of the VM migrations. "As we make this infrastructure more dynamic, it becomes more challenging," he said. "Having some pieces of software that reach down and automate this is a direct challenge not just in terms of infrastructure but in terms of the organization that manages it. Virtualization 1.0 was about consolidating multiple servers into one box. Virtualization 2.0 is around orchestrating multiple boxes into that virtual layer, and a whole bunch of virtualized functions get consolidated into that layer."
The movement of VMs can make it difficult to know which physical ports are being used by which applications. With physical servers, it was possible to simply make a spreadsheet noting these relatively static relationships, but as VMs move, virtual machine management tools help automate the tracking of these relationships. Some variations of these tools include HP's Opsware suite, VMware's VirtualCenter 2, ToutVirtual's Virtual IQ, Platform Computing's Symphony, BMC Software's Virtualizer, and BladeLogic's Operations Manager.
Molly Stamos, Opsware Group Product Manager at HP, said: "If I have a port flopping issue where the interface is going down, I need to know what applications are behind the switch. When you bring virtualization into the mix, it becomes harder to figure out. Now it is not just the physical server at the other end of the port, it is the physical server and the 10 VMs running on that port. There is no spreadsheet in the world that can keep that up to date. They need a way to discover that topology in real time."
Coordinating with virtualized server management
One of the greatest benefits of a virtualized infrastructure is also one of its greatest challenges. The infrastructure makes it far easier to move provisioning down to a wider IT community in which more individuals can set up a new server in a matter of minutes, compared with the weeks it takes with physical servers. These power users can create their own applications quickly with all of the requisite expirations and charge backs built in.
They also need to set up the virtual networking switches to connect to these virtualized applications, Stamos noted. "Virtualization is even further blurring the lines between the network and the server," she explained. "A lot of times, application developers will configure these virtual network devices without knowing a lot about networking, and then the network guys will have to come in and troubleshoot the problem."
With VMware's ESX, for example, it is possible to set up 32 virtual switches, which can be segmented into VLANs. Stamos said that server administrators sometimes inappropriately configure the switches to create a loop. "A lot of times, the server administrators don't understand VLANs and VLAN Trunking Protocol," she noted.
When someone makes a change in the network, it can cause an issue. The individual making the changes may not realize what impact it will have. Tools like Opsware help an organization define best practices for networking configuration settings and notify both the offender and the networking management team when changes violate these practices.
"Customers are going to have to track changes, configuration and performance management of VM systems," Elliot explained. "It has to be budgeted and planned for. If not, the risk of application availability and systems error put the system at risk. How this is going to be done is a work in progress."
George Lawton is a freelance writer, based in San Francisco, who has written more than 2,000 stories for SearchWinDev.com, IEEE Computer, and Wired (among others) over the last 17 years. Before that, he helped build Biosphere II, worked on a cattle ranch in Australia, and helped sail a Chinese junk to Antarctica. You can read more about him at his website, www.glawton.com.