As enterprises become more involved in virtualizing their servers, they are finding that virtualization can deliver more than just better utilization of their computing resources. But the potential risk for downtime can increase substantially.
The value of one particular physical server goes up as it carries more and more virtual machines (VMs) running on it. This makes having a failover solution more important.
“If you lose one of these servers entirely, you could have a problem, particularly since on average, eight or more VMs are typically running on a single box,” says Carl Drisko, a Novell executive who specializes in virtualization.
As virtualization becomes more prevalent, features such as high availability (HA), live migration of running VMs and near-term server failover – previously only the province of very expensive clustered configurations – are now available in the virtual world.
But having a truly HA VM world isn’t so simple. There are three factors that are required to roll this out in your data center:
1. Tools that are part of the VM hypervisor or work closely with it, such as live migration and orchestration software that can bring up new VM instances when existing ones fail or need to be brought down for maintenance.
2. HA-oriented storage networks with redundant arrays that can failover to each other and can handle continuous data protection, such as Datacore Software offering or arrays from EMC and Dell.
3. Backup software that can handle granular file restore from VMs.
Let’s look at some of the recommended products in each category. We summarize them in the accompanying chart below.
First and foremost is VMware. While they are the leading VM vendor, they have an ever-changing and somewhat confusing collection of tools as part of their vCenter management offerings. Offerings include the vCenter Server Heartbeat, their Lifecycle Manager and vMotion for Live Migration. The downside to using their tools is that they are expensive, as you pile on the add-ons.
VMware also has the largest third-party support for HA add-ons, including products such as Steeleye LifeKeeper, Double-Take Availability, and Veritas Cluster Server. All can monitor both physical and virtual applications, among other products.
If a host of the VM dies, or if the application hangs inside the VM, they can automatically bring up another host and a working VM, in some cases with the same IP address to continue the application. These products can typically monitor the service level of the application, and whether they are on a physical or a virtual machine. It is worth investing in these and similar third-party tools for their flexibility and the failure situations they can protect against.
Leading competitor Citrix has its Essentials for XenServer, which includes a variety of tools, including high availability, to manage both its own XenServer and Microsoft’s HyperV hypervisors. For example, it includes a fully automated system failover for its VMs as well as workload orchestration. A few of the above third-party tools also support XenServer, too.
Microsoft’s high availability features for its Hyper-V hypervisor are included as part of the latest version of Windows Server 2008 R2, including live migration and better support for storage area networks. Microsoft also has a companion tool called System Center Data Protection Manager SP1 that supports live host level HyperV backups and bare metal restores too.
There are other tools that can orchestrate bringing up new VM instances when workloads change, such as Novell’s Platespin series of tools – while these aren’t strictly for HA situations, they can contribute to having higher availability servers.
In some situations, having a virtual server go down for a few minutes while a replicant is brought up is a sufficient service level for non real-time applications, such as email.
Moving on to storage networks, most of the major players such as EMC, Hitachi, IBM and Dell have a variety of HA-tools to create and synchronize redundant storage arrays that can be used by VMs.
Alternatives to using the hardware suppliers are companies such as Datacore Software’s SANsymphony and Symantec’s Veritas Volume Replicator
that include remote replication and continuous data protection features. CA’s XOsoft has HA and storage replication options that can work with all big three VM hypervisor vendors (Microsoft, VMware and Citrix) – although there are different features supported for each one.
Finally, don’t forget about backups and making sure they are HA- and VM-aware. For example, Symantec Backup Exec 2010— has agents for Microsoft and VMware servers that protect the entire hypervisor and all of its underlying guest OSs. You can do granular file recovery from deep inside the VM, and decrease recovery time, too, all with a single backup pass by the application. What this means if somehow a single file gets corrupted inside a VM, you don’t have to restore the entire virtual disk file image but just that particular file.
The software also works to support live migration and deduplicate data that is shared across multiple VMs, cutting down on storage requirements.
As you see, this is an emerging arena and there is still a lot of integration of various products with the VM hypervisor and backup vendors. The good news is that the third-party vendors are improving their support and the hypervisor vendors are putting more HA features into their basic packages.