A workload’s performance is effected by many factors. The following list includes some of the most common factors:
- Network bandwidth and latency
- Storage bandwidth and latency
- Memory size, memory speed
- Processor performance, number of processors or cores
- Application architecture and implementation
Encapsulating a workload using either virtual machine or application virtualization technology could serve to exacerbate existing performance issues. The use of this very same technology can allow organizations to overcome some important application limitations including number of processors supported or the size of memory the application can address. If the virtual machine is moved from one place to another, the factors of network and storage bandwidth could become a significant factor as well.
In many circumstances the overhead imposed by the use of either virtual machine or application virtualization technology may be totally submerged by the limitations in network, storage or memory subsystems and thus, not be an important factor.
Richard discussed many improvements that are currently available in VMware’s ESX Server 3 and the impact they’ve had on performance. This product is much better than its predecessor. The overhead imposed by the hyperviser could be considered negligible in many applications today. VMware clearly intends to further reduce the “virtualization tax” its products produce in the future. As I”m able to say more, I’ll address this issue again.
Richard also demonstrated some amazing performance gains when virtual machine technology was used to overcome the processor or memory constraints built into some popular applications. Once again, as I’m able to say more, I’ll post something more specific.
VMware is clearly focused on busting the myth of the virtualization tax.
If virtual machine software’s overhead was reduced to somewhere below 10%, would your organization use this technology more broadly?