Implementing disaster recovery (DR) solutions has been one of the most difficult tasks for IT departments over the past several decades. So, it's not surprising that in a typical enterprise, only the most mission-critical applications are protected via remote replication to a DR site. In a typical small to midsized business (SMB), even fewer applications are protected in this manner.
We estimate less than 20% of applications are protected by DR in a typical enterprise and the number is smaller for a typical SMB. However, all of that is changing due to server virtualization's trumpeted march into IT shops of all sizes and its inherent ability to simplify DR.
DR in the physical server world has required that identical servers be available on both sides. In a large majority of the cases, even storage has to be identical. That means every aspect of the physical server on the primary site has to be identical to the one on the secondary site. This includes the server model, revs for the OS, BIOS levels, Windows registry settings, HBA and NIC models and driver levels. The same goes for application and security revs, as everything must always be identical.
The expense associated with buying and maintaining identical pieces of gear on both sides has been too much for most organizations. Out of necessity, DR is relegated to only those applications whose value to the company is unquestionable and company viability would be jeopardized without such an application being available. Financial data, customer orders, inventory and manufacturing status fall in this category.
I can't count the number of times I've heard a customer say, "I don't mind buying and using an EMC Corp. Symmetrix for the primary site, but why can't I use an EMC Clariion for the secondary site. I can deal with lower performance in case of a disaster as long as the systems are available." The same scenario is true for storage from other vendors.
Also, testing DR is disruptive to production, the process of recovery is manually intensive, and there is no assurance that all systems are functional or that recovery will indeed occur. The visibility into DR processes has been so poor that a new category of software called disaster recovery management, or DRM, has appeared on the scene recently simply for this purpose.
The benefits of server virtualization
Enter server virtualization. By abstracting away the physical server components and representing everything in logical terms, server virtualization eliminates one of the biggest issues of DR: the need for exactly the same equipment on both sites. It allows all system elements, including server make/model, OS release and patch levels and make/model/driver levels for HBAs, NICs and storage to be dissimilar, without jeopardizing recovery. Similarly, one is free to choose the style of replication, be it host-based, storage-based or network-based.
Additionally, it simplifies the recovery process by encapsulating not only the data (as done in a physical DR situation) but the entire system, including system configuration, OS, data and applications into a simple file(s), which can be easily replicated to the remote site. Recovery is simply running this file as a virtual machine in an appropriately configured physical server on the remote site.
By replicating this file on a periodic basis to the remote site, the recovery point objective (RPO) is improved. The fact that recovery is almost instantaneous means the recovery time objective (RTO) is improved. But most importantly, the file can be run on the remote site anytime to test the viability of DR without impacting production on the primary site. Given the fact that systems on the remote site can be "lesser" than those on the primary site and can be of different makes/models, the IT organization can use existing equipment without worrying about rev, patch and security levels. This is a major step forward for DR.
Lower costs means IT can afford more applications to be protected in this fashion. Easier implementation and essentially hands-off management means IT can broaden the number of applications without the prospect of ulcers. Server virtualization is a paradigm-shifting development for DR.
According to VMware, more than 55% of users are getting into server virtualization to obtain DR benefits. In other words, DR has become a major driver for server virtualization. Anecdotally, we find that one of two IT folks we talk to thinks DR is one of the two main reasons for introducing server virtualization (the other being server consolidation). Server virtualization is bringing DR to the forefront and making it easy for the masses to enjoy the benefits of remote replication and DR.
This could not be happening at a better time. The post-Sept. 11 world demands protection from site disaster. But so does the reality of power failures, floods, earthquakes and other natural disasters, not counting human errors. The fact that DR had been reserved for "elite applications" had nothing to do with the business' need for protection across a broader front. Rather, it had to do with the cost and difficulty of doing DR.
All this has changed with the advent of server virtualization. Server virtualization is in its infancy with less than 5% of the worldwide servers virtualized. The need for protecting large numbers of applications from site disasters couldn't be greater. Technology has finally caught up with the needs of business and the times we live in. Developing and implementing a DR plan is now more feasible and affordable. The era of broad-based DR is upon us -- and not a moment too soon.
About this author: Arun Taneja is the founder and senior analyst of the Taneja Group.