Handling virtual and physical servers together
One solution is to lease machines that you are providing, such as a server blade inside a blade system. This might not work for all customers who might want just a single machine to deal with redundancy. In that case, suggest installing a server appliance such as the PlateSpin Forge (www.platespin.com/products/forge/). The Forge allows you to replicate multiple servers to a single appliance which stores the servers as virtual machines but can be configured to bring those VMs online when the master server fails.
The Forge appliance is a rebranded Dell server with two Quad Core Intel processors, 16GB of memory and a lot of hard disk space, running the VMWare ESX hypervisor. During the configuration you add the Forge appliance to the domain. Once this is complete you add workloads and set their level of protection.
A workload, in PlateSpin parlance, is a server – physical or virtual complete with OS, applications and data stack. It is replicated to the Forge appliance using Continuous Data Protection and the protection level you choose is applied. The advantage of this appliance is that you can use multiple levels of protection – snapshots, file-based replication or block-based replication. The latter is the most efficient, only sending the blocks that have changed on the servers and will ensure that the replicated copy is as up to date as it is possible to be.
Unlike clustering, this is not a full solution but more of a business continuity option. It replicates multiple servers to VMs in the one appliance and when you configure the failover options you have to set the amount of memory that each VM will be able to use. This means balancing it against the needs of the business and using it as part of a larger recovery solution.
One version of Forge protects up to 10 workloads – which would suit a small business with demanding IT. As the Forge appliance is a network-attached device you could also host it at your own site and use the 25-workload model to serve multiple customers.
Using virtual machines instead of physical servers improves the reliability of the environment generally. These can be easily copied from one machine to another and are relatively easy to back up. When used with the management tools from vendors such as VMware, Microsoft and Citrix, you can even migrate the VMs as they run short of resources.
To do high availability in this environment you can look at a product such as Vizioncore esxReplicator. This allows you to take VMs and have them kept constantly in sync with each other. Should the master machine fail then esxReplicator will switch to the redundant copy. The initial copying of the VM takes time, as you would expect, and if you are looking to replicate to a secondary location then you should allow plenty of time to complete this. Once created the constant updates take little bandwidth.
One of the advantages of this solution over the Forge appliance is that as the client upgrades their hardware, the VM replication just points at the new server. They can also just change the machine that the replication points to in order to spread the risk and ensure that there are enough resources available should they need to bring the servers online.
Neither PlateSpin Forge or Vizioncore esxReplicator are high availability solutions. They are designed for business continuity and failover. This won’t suit all customers but they’re less expensive than buying hardware to build complete high availability clusters, and both offer protection either on or off site.