Drawing the branch office bulls-eye

Branch Office Best Practices Newsletter By John Burke, Network World, 07/25/06

A key question that enterprises must answer when providing IT services to branch locations is, "How good is good enough?" That is, what should performance targets be for enterprise application users across all locations? There are five basic strategies IT managers can follow:

1. Fine-grain, data-driven: Here, IT execs conduct detailed evaluations of each site's unique requirements, centered on business processes, the applications they depend on, and the applications' performance profiles. Performance profiling emphasizes the environmental factors (geography, WAN connectivity) and load factors (time of day and year, transaction patterns) that most affect critical transaction times and response times. The end result is an ability to instrument the environment to report on those key metrics, and to set meaningful SLAs around particular processes.

Needless to say, this is a rare approach. It requires a lot of work and a longish time to get to the point where the desired information has been collected, analyzed and used to formulate an SLA.

2. Coarse-grain, data-driven: Somewhat more common, enterprises use less detailed evaluations with analysis largely limited to the criticality of the site and connectivity to it: criticality dictates the minimum level of performance to which IT must commit; connectivity dictates the maximum level of performance to which IT can commit. Enterprises divide the results into tiers defining a few basic categories of site, and set performance targets for each category.

3. Internal market driven: Here, rather than defining the connectivity a site needs, or can expect, the central IT organization defines various performance levels it can provide - and the associated costs - and then lets the business lines pay for what they deem necessary. Essentially, this makes the whole thing not just a business decision but a business-line decision.

Altogether, these three strategies guide the efforts of fewer than half of the enterprises I work with. The rest implement one of two much simpler performance-management strategies.

4. One size fits most: This involves taking the slowest sanctioned connection - T-1, 56Kbps frame relay, whatever - and making the performance of all critical applications over that the baseline, promised performance for the entire enterprise. Of course, a requirement set that low is easily and consistently exceeded and meeting SLAs largely becomes a matter of ensuring availability.

5. We give you a wheel, tell us if it squeaks: This most common strategy consists of not making any commitments beyond connectivity and availability - not setting any targets for performance and accepting implicitly that it will vary depending on location and load - and simply responding to complaints as they arise. Altogether, nearly half of enterprises in our current research manage performance this way, and many admit that they over-provision connectivity as a result.

Which approach is best? Nemertes is currently conducting a quantitative assessment of success factors and best practices for leading-edge firms. Early indications are that the market-driven approach strikes the right balance between proactively addressing the real variation in needs for performance and doing so in a way that takes neither too much time nor too much effort on the part of IT.