Oli Thordarson knows that technology works best when the level of IT performance is directly linked to business goals. But many of his clients don't.
Thordarson, the CEO of Alvaka Networks, a managed service provider in Huntington Beach, Calif., tells of meeting with the CFO of one client company, a firm with approximately $40 million in annual revenue. "They were having problems with their infrastructure, and we identified their Exchange server as the biggest problem," he says. "We determined that, when the server went down, it cost the company about $50,000 a day. And with four to five outages last year, the company was feeling the dent. They had about four or five outages within the last year, which was a lot of dollars for a $40 million company." When the CFO realized the fiscal impact of these ongoing infrastructural glitches, the importance of managing his IT infrastructure performance suddenly became clear.
And that's where service-level management (SLM) comes in. SLM is the process of ensuring that IT services are properly linked to business processes and goals, and that there's performance criteria is in place to support those goals. "The fullest definition is understanding the service-level requirements of the customer and being able to translate that into a set of measurable metrics," says Jeff Kaplan, the managing director of ThinkStrategies, a consulting and research firm in Wellesley, Mass. "It may sound redundant, but some metrics are more easily measured than others, and most organizations forget to tie the technology metrics to the right business metrics."
SLM grew out of the outsourcing craze of the early '90s, Kaplan says. "IT organizations were trying to do better at managing vendors," he says. The movement started with service-level agreements (SLAs), which were contracts between service providers and customers that specified what services would be given and provided measurements to make sure that the correct level of service would be met.
Over time, IT departments at big corporations adopted the practice, writing SLAs for different business units within the company. When SLAs became difficult to measure and monitor, IT executives began putting in place tools and management techniques to better link SLA metrics with business goals, and SLM was born.
Today's cost-cutting emphasis has caused some companies to think twice about investing in SLM tools and expertise, Kaplan says. "Cost-justifying management software and adding on management expertise is seen as overhead," he explains.
Sadly, postponing SLM is a classic "cutting off one's nose to spite one's face" kind of move, since a good SLM program will save money by streamlining and better targeting infrastructure resources. Smart CIOs will take the long-term view and look into setting up an SLM program now. The following best practices can help:
Know your IT department's capabilities
There's no use creating an SLA if you have no idea whether your IT organization can deliver on it. Therefore, IT executives should start with a thorough inventory of their technical capabilities, making sure that all programs interact well together, including outsourced programs, says Jerri Ledford, author of Web Services: Understanding Service Level Management.
Malcolm Fry, an independent analyst and longtime industry expert on service management technology, suggests that CIOs create what he calls a "service catalog." "List all of the services you offer and the default values and parameters of each," he says. "For example, the default on Network A is 99%." (Of course, it's vital to have all the operational agreements and contracts in place to support the catalog.)
The catalog should then serve as a baseline for service discussions with business unit personnel. "The SLA is essentially the difference between what the customer wants or needs and the default service already offered," Fry says. "If there's a difference, that's when you start talking about how to finance the support and personnel that will be needed to provide the higher level of service."
Convene a team to run SLM
SLM isn't a one-shot deal. Proper implementation means that IT folks will talk to customers on an ongoing basis, checking performance and ensuring that the right things are still being measured. "You need to gather a team that's responsible for talking to the customer on an ongoing basis," Fry says. "You can't just say, 'Here's your agreement; see you next year.' This changes too quickly for that."
At the very least, the CIO should be responsible for getting the team up and running, says Ledford. "It's very important that it be handled at the top level. It's not something you want to throw off onto somebody else," Ledford says. Beyond the CIO, she likes to see IT managers, business unit managers and representatives from any outsourcers on board.
Invest in tools
Availability, throughput, response time, response policies -- all of these common service-level parameters need to be measured, and that means performance measurement tools.
"There have to be a variety of technologies put in place to monitor performance," says Kaplan. "In fact, there's been a whole industry built around this. BMC, HP, Tivoli and CA are all involved in this space." For example, Alvaka's Thordarson relies on Patrol Express from BMC to monitor the performance and availability of servers, applications and storage and network devices.
Measure the right things
For SLM to work properly, IT performance measurement needs to be linked to vital business goals. Therefore, it's important to make sure that the metrics accurately reflect those goals.
"IS as a rule has traditionally measured speed of performance and not quality of performance," points out Fry. "Many IT metrics show that they are doing a fast job, rather than a good job." For example, the classic help desk measurement of problems resolved on the first call doesn't encourage help desk personnel to actually improve quality -- that is, lower the total number of calls made in the first place. "Why should the HD want to improve quality?" asks Fry. "They'll get nothing for actually reducing the number of incidents."
Monitor and adjust constantly
Once the SLAs are in place, it's important to compile and analyze performance data and report it regularly to the SLA customer. All that should lead to a service improvement program, Fry says. "You want to know where you missed targets and how do you fix it -- buy more capacity, for example. It should be a continual improvement program."
And by continually improving the tie between IT performance and business goals, CIOs stand a better chance of building an IT organization that truly serves business success. Thordarson says that his company sends a questionnaire out to new clients that contains the question, "What issues and concerns weigh heavily on the mind of the CEO?" "If they can't answer that, they've got trouble brewing," he says.