|
Server virtualization is becoming more and more popular for disaster recovery (DR). However, server virtualization for DR has several pros and cons.
Pros of server virtualization for disaster recovery:
- Golden image: A golden image can be created as a corporate standard build. To deploy new servers, server deployment is as easy as cloning this image.
- Reduced disaster recovery costs and maintenance: For every physical server bought, your cost is doubled because an identical server also needs to be purchased for your DR environment. Also, when you buy physical servers, they need close monitoring and servicing at the hardware and software level, which can become pricey. But with server virtualization, hardware monitoring is reduced and capital expenditure is cut in half.
- Green IT: If you have fewer physical servers, you will use less power in your IT environment, which in turn will give you a greener IT environment.
- Storage technology: Since server virtualization is based on files, all the technology that is available to unorganized data is available to server virtualization. Snapshots, cloning, thin provisioning, deduplication, backup and replication are all available to your virtualization environment.
Cons of server virtualization for disaster recovery:
- Synchronous storage replication: Although the cost of disaster recovery decreases with fewer physical servers, the cost of synchronous storage array replication is expensive. Maintaining the infrastructure to support synchronous replication costs more than people would expect with the upkeep of long haul fibers and possible Fibre channel (FC) infrastructure.
- Asynchronous storage replication: The cost of asynchronous replication is much less than synchronized replication, however, to capture an image of your environment properly, you will need software that quiesces the environment long enough to take a consistent snapshot and replicate based on that snapshot. How often this is done dictates your recovery point objective (RPO). Whether to choose synchronous or asynchronous replication is determined by your goals, cost and what the business can tolerate as being acceptable.
- Limited resources on virtualization platforms: Since all virtual machines (VMs) run on a virtualization platform, all computing cycles are timesliced among all of the VMs running on that platform. With this model, heavy computing environments such as database servers are not conducive to virtualization. That does not mean that "real" servers should not be put on virtualization platforms. DNS servers, NIS servers as well as Active Directory servers are important servers yet their workload is not as intense as some other servers, such as a Sybase server. Infrastructure servers may be great candidates to server virtualization.
Overall, my recommendation is to test your server virtualization and test products before you implement them in your environment. It is more expensive to redo an implementation than to slow down and do it right the first time.
|
