| August 18, 1997 |
Good Service Guaranteed
Centralization and outsourcing spur interest in service-level management
By Lenny Liebmann
These are questions that service-level agreements between IS and business units are supposed to answer. SLAs aren't new, but several factors are making them an increasingly hot topic. Some organizations trying to centralize system management and other IS services need to show user departments that they're delivering service as promised. At other companies, the possibility of outsourcing is driving organizations to more clearly define the level of servic
e IS provides.
SLAs are becoming more viable because of the growing number of tools to monitor distributed computing environments. "SLAs are hot now because vendors have figured out that users will pay for tools that can monitor service-level metrics, and they've introduced those tools in the last six months," says David Passmore, president of Decisys Inc., a consulting firm in Sterling, Va.
Hal Uygur, VP of enterprise technology oper
ations at Goldman, Sachs & Co., heads an effort to centralize management of distributed systems at the New York financial-services firm. His company is working on service-level reports, built with Netcool event-management software from Micromuse in San Francisco, that will document his central tech-support team's response to alerts and verify that the team initiated action within two minutes in each case. "We need to show our internal customers that we can, in fact, move forward with centralized management," he says.
Centralization also drove the development of SLAs at the state of Michigan's Consolidated Network Operations Center in Lansing, which services 30,000 users in state agencies. "We needed to show the agencies they were getting credible service in light of the fact that we had taken things over from them," says Brian Ruhf, the Center's manager.
Ruhf sees the implementation of SLAs progressing through three stages. When the Center was formed two years ago, its SLAs simply defined how agenci
es would do business with the Center-who their contacts would be and how problems would be resolved. Now, the Center has moved on to SLAs that define commitments to service-response levels, based on the severity of the event and the importance of the service affected.
For example, a "red" condition-an essential device or circuit is down-requires a first response within 15 minutes, but a less-critical "yellow" condition requires a response only within a day. Users can ask for quicker turnaround, but must pay accordingly. To track and report compliance, Ruhf's group uses Remedy Corp.'s help-desk software and Spectrum, a network-management platform from Cabletron Systems Inc. in Nashua, N.H.
The next step is to monitor the performance of the computing environment, including link failures and bandwidth utilization. But Ruhf says few users are asking for that data. "They're much more interested in how fast we get moving when there is a problem," he says, "and how they can make sure we're doing what we te
ll them we're doing."
Some companies, though, find that there's strong demand for reporting application performance in SLAs-particularly when information is provided in terms users understand. With the growing use of standard business packages such as SAP's R/3, companies are more likely to find a tool to do the job.
Stephen Enk, director of IS operations for Molex Inc., an electronics manufacturer in Lisle, Ill., turned to Luminate, from Luminate Software Inc. in Redwood City, Calif. The tool documents how long it takes business users to execute specific tasks, providing Enk with a reporting tool that delivers meaningful information for SLA compliance. "If you tell business users, `The response time for transaction T0-53 is two seconds,' they're going to go, `Huh?'" Enk says. "It's much more useful to tell them how much time it's taking users to execute a task on the order-entry system."
Big enterprise-management vendors offer less-specialized tools that may help support SLAs. Hewlett-Packard rec
ently acquired a set of system-management tools, including products for monitoring application service levels. Tivoli Systems' TME and Computer Associates' Unicenter provide features such as business process views and application management. "Conceptually, CA Unicenter's business process views provide a good foundation for SLAs," says Paul Mason, a senior analyst at International Data Corp. in Framingham, Mass. "But none of the enterprise-management vendors have really delivered anything yet."
The use of SLAs based on performance data is still limited. Of 30 technology managers interviewed for this story, about half use some kind of SLA, and eight more are developing SLAs-but only four provide internal departments with regular reports on the performance of networked applications as part of an SLA. A survey by the Enterprise Management Institute in Boulder, Colo., found that just 13% of IT departments use data from management applications to generate SLA compliance reports.
IT managers who do want to
measure application performance often must create their own solutions. "We use a combination of the statistics we can pull from our Unix servers-CPU busy, CPU wait, run queues, etc.-and scripts we create that actually generate application transactions, so we can measure end-to-end response times," says Steve Rees, computer services director at hardware and software vendor Tektronix Inc. in Beaverton, Ore.
Denis Haensler, a network manager at pharmaceutical company Hoffman-LaRoche's Basel, Switzerland, headquarters, uses InfoVista from InfoVista Corp. in Redwood City, Calif., to pull together network-management data into service-level reports. InfoVista "allows me to bring together entirely different report types into a `dashboard' that nontechnical people can easily understand," he says. It also lets Haensler group information from multiple devices into a single view, he says.
SLAs are often linked to payment for services, and that relationship can become more complex as tools evolve that can monitor
performance. "Once we get systems-performance issues squared away, we're planning to put in provisions for credits in case we don't meet expectations," says Steve Lewack, network services manager at Alverno Administrative Services Inc. in Beach Grove, Ill., which runs hospitals and health-care facilities. "That also means we'd like to be able to have incentive bonuses for our people if they exceed their goals." Lewack's department has committed to providing 95% network uptime-downtime is defined by user minutes of downtime rather than network link failures.
Outsourcing Threat
Another situation that drives the development of SLAs is when internal IT operations are broken out as ind
ependent units that compete for the business of other units in the company. "SLAs are essential in our situation," says Richard Schmidt, manager of systems management for Shell Services Co., a U.S. Shell unit in Houston. "They give us a target that acts as both a driver for internal improvement and a basis for establishing credibility with customers."
But not everyone sees SLAs as the basis for effective working relationships. "If you have an outage, it doesn't do you any good to point to an SLA and say, `We only guaranteed you 99.5% availability, so we're entitled to this,'" says Michael Smith, director of network strategy at Home Box Office in New York.
Yet vendors and analysts agree that there's increasing demand from business units and IS groups for SLAs and tools that help implement them. "Business units want to give up management because it's become way too complex for them," says Luminate president Bruce Fram. "They need to know the service they're getting-whether they go to an outside vendor
or use the corporate IS department."
|