8 FEBRUARY, 2005

Is the SLA dead?

A passing remark by David Gee, worldwide VP of software marketing for HP's Technology Solutions Group, has tended to confirm a thought that has been rumbling round my mind for a little while now - namely that IT's favourite management tool, the Service Level Agreement (SLA), may be measuring the wrong things for the wrong reasons.

Gee's remark, that SLAs are a 'Babel Fish issue' (in a neat piece of re-use of the well-known translation `tool' from `Hitchhiker's Guide to the Galaxy'), highlights what is likely to become a growing problem. Why he thinks the dreaded fish is needed is an old problem: the ability of business users to speak `IT' and IT staff to speak `business' remains desperately poor, and what is needed is something that will span the chasm that seems to exist between the two.

What lies at the heart of this, in the end, is that IT tends to define itself and its operations in technical terms. Systems have 99% uptime, servers respond in 1.34 seconds or less, and so on. These are very good measurements of pure system performance, but have as much relevance to a clerical assistant's job as the measurement of the consumption of ink in ball point pens has on their ability to write clearly.....the wrong measurement for the wrong reasons.

When it comes down to it, a system can be running within the thresholds set by contract in the SLAs - by definition it can be a 'good' system - yet if it is running within one or two points of breaching those thresholds it is actually running rather badly in business terms. It can be showing itself to be good as an IT system, but could still be losing the business 10 percent of sales a day. This would happen because the business processes would be running slowly enough to still be annoying to whoever is consuming it - be that an individual trying to start a mobile phone contract or a major corporate customer chasing urgent information on their million Dollar order. That can easily end up as a significant loss if it goes unnoticed for a few weeks.

This is now becoming an area where solutions architects need to think more cannily about what metrics they are going to specify for the systems they build, and why they are used. SLAs may be widely accepted and well understood, but maybe the real measurement of the system is the number of orders being taken, and the time taken to process them. Or perhaps it is where backlogs are developing in a supply chain. Once that problem is known in business terms, then and only then is it necessary to drill down into the detail of why the situation is happening so that remedial actions can be taken within the IT systems.