Implementing high availability in a virtual data center requires a bit of planning around your redundant network architecture and storage pools.
Many experts recommend installing as many network interface cards (NICs) as possible in your host servers and then dedicating a separate NIC to each virtual machine (VM). That way you don't have to worry about a virtual server depriving your other virtual servers of bandwidth, because it's running an especially bandwidth-intensive application.
Although this type of design works well for optimizing performance, it doesn't work very well for creating a high-availability architecture. Because our goal is to create a redundant network architecture, each host server will still require multiple NICs, but those NICs will be used in a different way.
Building a redundant network architecture
Ideally, each host server should have at least four NICs. You might consider using the Dell PowerEdge R900 rack server, or something similar, because it offers four built-in Ethernet ports. Currently, most of the servers on the market only offer two built-in Ethernet ports. Although many servers offer expansion slots for adding additional NICs, it is best to leave these slots free for other things like Fibre Channel cards, if possible.
Rather than dedicating each NIC to a specific VM, a better approach is to connect each NIC to a separate switch. Two of the switches should connect to the main network, and the other two switches should be used for backbone networks between your host servers. That way, if there were a switch failure, NIC failure or cable failure, each host server would still have a way to connect to the backbone network and to the primary network.
If your host servers are running VMware, then you would also have to configure the vSwitches for each server. A vSwitch is a virtual switch used to route traffic between the physical network and the virtual network.
This redundant network architecture ensures that virtual servers are not rendered inaccessible by the failure of a networking component. Although this type of architecture does provide network redundancy, it is not a viable high-availability strategy because the failure of server hardware still has the potential to cause an outage. A redundant network architecture is, however, a step toward creating a resilient virtual data center.
Putting VMs on a storage pool
A key task when creating a resilient data center is to centralize your storage. The idea is that your VMs should be placed on a centralized storage pool that is accessible to all of your host servers. That way, if a failure occurs, the virtual server can be moved to another host.
This type of resiliency would not be possible if the virtual hard drives were located on direct-attached storage because they would be tied to a single host server. The real trick to centralizing a storage pool is to do it in a way that does not create a single point of failure.
One way to accomplish that is to install multiple Fibre Channel cards into each host server and to connect each of the Fibre Channel cards to a separate Fibre Channel switch. The switches should all connect to a common storage pool. The storage pool itself should consist of redundant arrays.
In part 3 of this tip, learn how live migration and fault-tolerant systems can further boost business continuity.
|Brien M. Posey, MCSE, has received Microsoft's Most Valuable Professional Award seven times for his work with Windows Server, IIS and Exchange Server. He has served as the CIO for a nationwide chain of hospitals and healthcare facilities, and was once a network administrator for Fort Knox. You can visit his personal website at www.brienposey.com.|