Virtualization stall usually occurs in a server virtualization project when business-important or mission-critical servers are targeted for virtualization. There is risk associated with virtualizing these servers and this risk can cause IT managers to move slowly or put off their conversion. Part of the problem is that once a server is committed to the virtual infrastructure the ability to ‘back out’ to the physical world is difficult. Using replication to test the load of a future virtual machine and to provide an out in case something goes wrong is an effective way to reduce this risk and eliminate virtualization stall.
What is Virtualization Stall
Server virtualization projects, no matter whose data center they are in, tend to follow the same cycle. They start out in a test phase, which often goes well, and then move quickly into basic production, which includes those applications not considered ‘mission critical’ or even ‘business-important’. When things go well here, server virtualization has essentially proven itself and the next step should be full production. This is the phase where business-important applications are virtualized and possibly mission-critical ones as well. It is at this point where virtualization stall tends to occur.
As mentioned above there is risk involved. There are concerns not only about whether the business-critical application will work in a virtualized environment but also what the resource impact of virtualization will be on that environment. Unlike the physical world, virtual server platforms overcommit resources, which can often lead to problems if not managed effectively. The concern is whether virtualizing the servers which run a set of business-important applications will result in performance degradation for those VMs or cause a problem elsewhere in the virtual environment.
Replication solutions like DoubleTake from Vision Solutions can help prevent virtualization stall by providing a means to test the load of a future virtual machine, provide the ability to perform that switch to virtual mode in a controlled fashion and to provide an ‘out’ if something goes wrong. These products use a process called “change-based replication” to maintain data sets on two identical servers – referred to as the “source” and “target”. Typically these software solutions capture disk writes on the host server and send writes to a target server. Then they replicate just these changes between the source and target servers, keeping both data sets in synch.
They run at the filter level of the source machine and have no awareness of the application or data that is being synchronized. This byte-level replication can be run continuously, creating a target that’s updated essentially in real time. This process works as well in the virtual environment as the physical running on the guest OS so that changes can be captured in real time and the target can be another virtual machine or even a physical host.
Replication as a load test
In virtualized server environments, there’s always the fear of overloading a host server with too many VMs. Load testing would seem to be the common response, although for most virtual environments, this would require time and resources to accomplish. Replication software solutions can provide a simple and effective way to test a host configuration to see if it can support critical VMs.
There are a number of scenarios for load testing, but one involves setting up a test VM and synchronizing it with the production server that’s a candidate for virtualization. To start the process, the synchronization between these two is setup and the physical server is replicated to the VM. This process creates real time duplicates of these physical servers on the test VM, without impact to the physical. If everything runs well, after an adequate time period, the original physical can be migrated to the virtual host and the job’s done.
At some point the replication can be reversed and these now-critical VMs can become the primary and the original application server the target to make sure the application can be sustained by the virtual environment. When this has been confirmed, the physical servers can be shut down and the production applications run on the virtual servers alone. A final step may be to replicate the new virtual machine to another virtual machine on a separate storage system and virtual host for added reliability.
Replication as an Out
In both of these scenarios, replication can provide an out, or a ‘do over’ for the transition of business-important or mission-critical applications to the virtual server environment. If the result is less than satisfactory, the VM can be shut down and the physical server will continue as before. Products like DoubleTake can provide a simple, redundant infrastructure to set up and run for any number of physical servers which are slated for virtualization. These source – target pairs can be any combination of physical or virtual servers.
This original application server running in parallel with the now production VM, can provide the ‘back button’ should something unforeseen occur to any of the VMs or applications. This allows for both performance and support issues to be addressed where virtualization itself may be to blame. At some point in the future, this physical stand-by may be decommissioned, or it can be used as a disaster recovery solution, locally or off-site. It can also be the “global” physical server where multiple VM’s replicate to it. In this case any virtual machine can quickly become a physical stand alone system.
Lower cost risk Control
All risk can never be eliminated and the cost, in preparation and testing, to reduce the perceived risk prior to virtualizing a set of business-important servers may be quite high, depending on the scenario. In contrast, the ‘undo button’ capability provided by one of these replication solutions may eliminate a large portion of that preparation and testing. The result could be a significant cost savings which improves the return on investment of the virtualization project itself.
Virtualizing servers, like any change to an IT infrastructure, brings a certain amount of risk. This risk can cause virtualization projects to stall, especially when having to virtualize the servers running business-important applications. This delay can impact the ROI of virtualization projects and increase opportunity costs.
Replication solutions, like DoubleTake, are an effective way to reduce the risk of server virtualization. This technology enables the creation of a secondary host server infrastructure without disrupting the operation of the primary host. This secondary host can be used to load test a set of VMs or confirm their impact on shared resources before committing the new infrastructure to production. By providing a way to effectively ‘undo’ the physical-to-virtual conversion if something goes wrong with a business-critical VM, these solutions can reduce the mitigation steps often taken and eliminate virtualization stall in the process.