|
Virtual machines (VMs) often require the same types of resources as their physical counterparts. For example, a Web server workload will generally require the same amount of bandwidth whether it's running on a physical server or a virtual machine. That's not so bad, overall – most organizations are already looking at capacity planning and performance monitoring for their current physical servers and extending that to VMs isn't a huge stretch.
The major issue to keep in mind is saturating or overload physical host server resources. You need to plan for the aggregate load of each VM, add the load placed by any applications or services running directly on the host, and then include a virtualization "overhead" factor. Important subsystems include CPU, memory, disk, and network resources. For example, if you have a single Gigabit Ethernet physical connection on a server and you place a bunch of network-intensive VMs on that machine, the physical NIC might quickly become a bottleneck. You'll need to look into using multiple NIC ports (which is supported by most virtualization platforms) and features like load-balancing and NIC teaming. Overall, the issues can be addressed, but don't get caught off-guard by failing to do some performance profiling before going into production!
|
