Delivering high availability (HA) to applications classified as “business critical” in recent years has been as much a technical obstacle as a financial one that organizations have struggled to overcome. The latest version of Symantec’s Veritas Cluster Server addresses these concerns. Now any application running on either a physical or virtual machine may recover almost immediately to a virtual machine (VM) giving enterprises the high availability (HA) they have sought without the hardware costs or VM reboot wait times.
One of the primary motivations behind enterprises wanting to put business critical applications in HA configurations as they virtualize them is simple: they want to minimize or ideally eliminate the possibility of downtime once they are virtualized. It is for this reason that many enterprises initially turn to HA solutions provided by VMware to address these concerns.
VMware offers HA, vMotion and Site Recovery Manager (SRM) as well as utilities such as its Distributed Resource Scheduler (DRS) that aligns available compute resources with the business priorities of the applications running on individual VMs. However they do not account for all ways in which an application may fail. While the native HA tools do an adequate job of managing the VMware environment, there are some considerations to take into account:
- They only work on VMware vSphere. More enterprises are moving to multi-hypervisor environments so enterprises need separate HA solutions for applications running on other hypervisors.
- They fail to account for business critical applications that remain on physical machines. Enterprises may need to continue to run some business critical applications on a combination of virtual and physical machines. VMware provides no technique to protect applications running on physical machines or the physical infrastructure from failures.
- Up to 25% of application failures are attributable to problems with the applications themselves. Symantec finds in its end-user surveys that up to 25% of application outages are directly attributable to a problem with the application itself. VMware provides no way to detect these specific types of application outages within VMs.
It is this set of challenges that the latest release of Veritas Cluster Server addresses. It has for years enabled organizations to put multiple physical servers into a cluster, configure one physical server as a standby server (all servers in the cluster may also run in active mode) and, enable any application to recover to that server in the event of an application, operating system or hardware outage or problem.
Differentiating Veritas Cluster Server from other HA solutions is its ability to detect application failures in addition to OS and hardware outages and then take action to recover the application on a healthy virtual machine. Veritas Cluster Server can do this without the requirement of a virtual machine reboot, resulting in less downtime.
While I covered a number of these new features at a high level in a prior blog entry, what I did not cover were some of the specifics as to how Veritas Cluster Server is configured to enable organizations to accomplish this without necessitating a VM reboot.
To do so, organizations need to first create an “extra” VM on the VMware ESX server that runs in passive mode. This VM now assumes the role of the hot physical standby server and hosts no applications while operating in standby mode. Then if an application (physical or virtual) fails and needs to recover, Veritas Cluster Server detects this event and simply brings the application up on the standby VM without needing to reboot it.
To configure the recovery of applications to this standby VM, organizations have one of two options. Veritas Cluster Server integrates with vCenter Server so administrators may use the vCenter console to configure applications running on other VMs to recover to this VM.
The main limitation is that vCenter may only be used on machines running VMware.
If organizations want visibility and control of their virtual and physical environments from a single console, they may use Veritas Operations Manager which is included with Veritas Cluster Server. (Additional features available in Veritas Operation Manager are covered in this blog entry.)
Business critical applications are many times and in most organizations every bit as critical as mission critical applications to their day-to-day operations. These business critical applications often just lack the financial arguments and appropriate HA software needed to safely virtualize them.
By Veritas Cluster Server now adding support for monitoring and failover of applications running on virtual machines,, organizations take these objections off the table. Now using a VM that runs in standby mode and consumes minimal or no server resources, organizations finally get the HA solution they need to recover their business critical applications without the hardware price tag and/or software limitations that accompany the deployment of native tools or other HA solutions.