Stuck with Legacy Virtualization Requirements?

Are you running applications from vendors who are cautious to adopt new virtualization functionality? This seems to be a common theme for Electronic Medical Record (EMR) Systems, which can create challenges for System Administrators.

With what seems like the never ending rise in the cost of Information Technology Systems, IT Staff including System Administrators must seek new ways to reduce the overall cost of a system. Some of these cost-saving opportunities come from license negotiations for software being used in a different area of the organization’s environment.

During a recent EMR system migration, the selection of the Hypervisor was driven by licensing negotiations. Microsoft made a proposal for Hyper-V that was acceptable to our customer for reducing the cost of their virtualization environment to host the EMR Database Servers. The customer’s EMR vendor supports Hyper-V, but the adoption rate of Hyper-V to host the database servers is low.

One of the challenges during this implementation was finding the best way to handle backups of the production OLTP database. For this particular EMR, the production database can consume 20 Terabytes or more of storage, which must be backed up on a nightly basis.

The OLTP database experiences random reads and eighty second write burst cycles. These write burst cycles require a limit to the amount of time that the database can be frozen during a backup cycle. The current maximum amount of time for freezing the database is nine minutes, and ideally, a database should be frozen for a much shorter period of time than nine minutes.

To accomplish a backup and to have the database frozen for no more than nine minutes, we typically configure array-based snapshots to meet the narrow timeframe. When hosting the database VMs on the VMware Hypervisor, raw device mappings (RDM) are used to facilitate the management of the Array Based Snapshots. RDMs make it easy to present them and manage the corresponding volume groups and filesystem at the VM level.

With the Hyper-V hypervisor, a pass-through disk is the closest thing to the concept of an RDM. Unlike a VMware RDM a Hyper-V pass-through disk will not move between physical hosts during a live migration of a Virtual Machine.

Because a pass-through device is not supported for live migration for this particular implementation, we used NPIV (N-Port ID Virtualization). Unfortunately, NPIV within Hyper-V has challenges when doing live migrations.

If I was to recommend a Hypervisor today for a workload with legacy requirements, I would recommend VMware. A big reason that VMware offers the support for legacy requirements is due to how long VMware has been in the Hypervisor business. Microsoft’s Hyper-V is comparatively new to the marketplace.