* Steps IT organizations must take to meet the key requirements of an SLA
The last newsletter discussed some of the political challenges associated with creating an internal service-level agreement for the performance of the company’s key applications and services. This newsletter will address some of the steps IT organizations must take in order to ensure that it is able to meet the key requirements of that SLA.
While not necessarily a requirement, we believe each SLA should state at least one cost metric. We strongly believe each application SLA must state at least one or two performance metrics. For example, the SLA might state that the application response time will never exceed 5 seconds. There are, of course, a lot of other ways to state an application response time goal. The SLA might say that the application response time will not exceed 5 seconds for any user in the U.S. and state different response time goals for international users. Alternatively, the SLA might state that the response time will be 5 seconds or less, 98% of the time.
Part of the challenge of meeting the requirements of an application-specific SLA is that any component of IT infrastructure could be the cause of application degradation. This includes the LAN; WAN; servers; operating systems; access points; WAN optimization controllers; firewalls; intrusion detection system; intrusion protection system; application delivery controllers; Web servers; application servers; database servers or storage. You get the idea – the typical IT infrastructure is complex and becoming more so every day due to factors such as the virtualization of basically every component of IT. This added complexity introduces a growing number of devices that could either fail or which could introduce additional latency into the end-to-end data flow.
We do not mean to diminish the difficulty of ensuring the availability of the IT infrastructure. However, the individual components of the infrastructure do not fail very often and when they do, there is often some kind of a backup, such as an alternative path through the WAN. The really tough challenge is managing how much delay each component of the IT infrastructure contributes to the overall application response time.
This is another situation where the IT organization has to face the fact that not all components of the IT infrastructure are equally important. In particular, while crafting the SLA for a given application, the IT organization must be able to identify those components of the IT infrastructure that support that application. The next couple of newsletters will describe why that task is so critical. In the meantime, we continue to seek your input about your experience with SLAs.
The next IT Roadmap conference will be held in Chicago on April 2. If you are in the area, try to attend as we will continue the discussion of how IT organizations can best cope with the current economic challenges.
