Principles of good virtualization

IBM supports virtualization and consequently, IBM Rational’s products are supported on virtualized servers. In a properly managed VM, Rational software products can perform properly. However, we insist that the virtualized infrastructure be properly managed and monitored. It is crucial to understand how your virtualized infrastructure uses affinity and overcommitment, and to be sure you are using affinity and overcommitment in a way that ensures the best performance of your IBM Rational products.

Affinity

Affinity (also called entitlement, pinning, dedication) is the ability to dedicate one or more resources on a virtual machine (e.g.: memory, processor, etc.) to the corresponding resources on the hypervisor. The host parcels out resources as the virtual machines need them. Affinity ensures that the wanted resources are dedicated to that virtual machine and are always available when the virtual machine requires them. Remember that virtual machines share system resources with all the other virtual machines on the same host.

Overcommitment

Overcommitment is when the total amount of virtual image resource allocation exceeds the physical resources of the hardware. In many common configurations to satisfy a virtual machine’s peak needs, the hypervisor may take resources from other virtual machines. Sometimes the combined needs of all the virtual machines may exceed the actual resources of the hypervisor. (When calculating the combined resources used, be sure to count the hypervisor resources as well). Sometimes over-commitment can cause all the virtual machines on a host to suffer.

Virtualization advantages

• Well-managed virtualized servers permit modifying CPU, RAM and disc image sizes with much greater ease than physical servers.
• A virtualized infrastructure can offer several capabilities of high-availability, such as the ability for the VM infrastructure to mirror standby servers in case of failure.

Recommended strategies

• Virtualization must be managed and monitored

• Ensure that CPU, memory and network resources are dedicated and uncapped; ensure that CPU, memory and network resources have minimum allocations or reservations

• For each hypervisor (VM host), the number of virtual CPUs should never exceed the number of physical CPUs; be aware of the trick played by some hypervisors which may count multi-threaded processors as separate processors (a dual-threaded 16-processor server may appear to the hypervisor as a 32-processor server and allocating 2 of these processors actually allocates half of a real processor)

• The CPU allocation for each VM corresponds to actual physical CPUs

• There is ample access (bandwidth, network cards) to network and storage

• If the server which is hosting your VMs (hypervisor or host) is not fully dedicated to your VMs, be aware of the usage and load of those other VMs

Related topics: How to determine whether a VMWare environment has a problem

External links:

• IBM developerWorks article Be smart with virtualization

Additional contributors: None