We’ve been using the Red Hat Enterprise Virtualization (RHEV) solution in our production environment at Red Hat for two months and I wanted to share our experience. We are a participant in the RHEV beta program and partnered with our product organization in support of the deployment.

We are operating 24 virtual machines in production. They include six JBoss cluster nodes for our Java services that drive www.redhat.com; four Knowledgebase cluster nodes used by our customers and customer support team; six mail exchanger nodes handling our internal email delivery; and four mail exchanger nodes handling our external email delivery. The physical infrastructure consists of a blade chassis with 14 HS21 blades (8GB RAM, dual core processor) and 1.2TB of SAN based storage provided by our Network Appliance infrastructure. The production nodes have been stable. They are performing equivalently to the bare metal nodes they replaced for both the JBoss and external mail exchanger workloads. Performance of the other nodes is very acceptable and in line with our expectations.

Also as expected, RHEV has given us the ability to monitor resource utilization and quickly provision new nodes. We have information and tools to plan for future capacity increases. The resource management tools are excellent. RHEV does a very good job of transparently managing storage, memory and processor utilization. The presentation of the allocation and utilization of resources is very clear. Live migrations, including automatic migrations, are very fast and reliable. The paravirtualized drivers are a significant improvement over base KVM and Xen and have performed very well. The management interface is intuitive and generally easy to use.

The RHEV hypervisor installs were easy and integrated well with our automated installation framework. As noted, the product is solid and resilient. We haven’t experienced any crashes or other hypervisor problems as are often encountered in beta environments. The virtual machines behave exactly as one would expect and respond to our existing tools just like our physical hosts. We encountered three issues during installation, but were able to resolve them quickly with workarounds or configuration changes. It took us roughly one week to install the machines and get them into production. Deployment has been a very solid experience for a beta environment.

In addition to our server deployment, we are working to deploy a RHEV cluster to provide desktop virtualization for end user applications and expect to have it up and running later this month.