Information Strategist: Rightsizing Data Protection

The current rage among vendors and press, at least when it comes to business-continuity planning, is the substitution of a high-availability architecture for conventional disaster-recovery (DR) techniques. CA's acquisition of XOSoft last year heralded the trend, which now finds advocates at EMC (having acquired VMware and Kashya) and elsewhere.

Using "active-active" or "active-passive" clusters to fail over from one site to another isn't a new concept. Companies with deep pockets and a desire for always-on operations have used this strategy for years. However, the price of the solution--which requires not only expensive hardware and software, but also a "hard" or predefined recovery site--typically has put it beyond the reach of less well-heeled organizations.

To hear vendors tell it, all that is changing. VMware's latest ESX Server as well as products from CA and Neverfail Group are moving HA to the fore. However, other factors are also contributing to the trend.

The state of Illinois implemented such a strategy in the wake of 9-11, for example, to work around the reluctance of legislators to spend money on anything called DR. When state IT managers asked for a disaster-recovery budget in 2002, they were rebuffed. However, when they instead explained that they just needed a place to test application changes, and pointed out that the state already owned a building some 50 miles from Springfield as well as the right-of-way between the sites for laying dark fiber, and that for a very small investment they could replicate gear and workload between the two sites, they were funded. Apparently, to get management buy-in on DR, you may just have to call it something else--anything else.

HA makes sense from a business-continuity perspective. It basically enables recovery and builds resiliency into IT architectures--a superior strategy in many ways to conventional disaster-recovery techniques, where the recoverability/resiliency of IT is simply bolted on.Three recent changes make HA more affordable: hardware commoditization (on the server side, at least); increasingly ubiquitous access to broadband networks; and standardization on Microsoft, Oracle and a handful of other OS and application software platforms. Previously, HA required identical, unused, replicated hardware at recovery sites. Today, those rules have changed: "generic" server environments can be duplicated without significant cost or work, and workloads can be balanced across multiple geographically separate sites to build resilient solutions that deliver improved operational business value.

However, replicating everything is needlessly expensive. The trick comes down to identifying business-critical processes, and their related infrastructure and data, and exposing the latter to appropriate recovery services based on a rational assessment of need.

While there tends to be a bias in many organizations toward the classification of all data from a specific application (usually an ERP or CRM database) as mission critical, all the data stored in volumes allocated to the finance or human-resource departments may not be equally critical--especially not all the Britney Spears videos someone has been downloading and storing on "critical volumes." Similarly, a lot of data may fall into the category of "never referenced, but retained for compliance" and may be better-suited for tape backup than synchronous mirroring.

Conversely, some data that may not seem important for recovery may be absolutely critical because it supports a mission-critical process, albeit indirectly. Only analysis and testing will reveal such interdependencies. For an example of an approach I have used to develop a better feel for data criticality, feel free to download a white paper at www.data institute.org/ dmb1.pdf.

HA clustering is the latest thing in DR, but it's just one approach in a spectrum of potential strategies and is by no means one-size-fits-all. Rightsizing DR to data is just as important as rightsizing storage to applications.Jon William Toigo is a CEO of storage consultancy Toigo Partners International, founder and chairman of the Data Management Institute, and author of 13 books. Write to him at [email protected].