Raise the bar on service-level management

Service-level management in the past took a “speeds and feeds” approach – service-level agreements contained specific service-level objectives centered around metrics such as performance and availability of certain infrastructure elements, such as a network link.

For example, a service provider might provide a guarantee of 99.5% availability for a frame-relay T-1 circuit. The problem with this is that it does nothing to guarantee that the users of that T-1 can do their work.

A few weeks ago I was negotiating a contract with one of our service providers. It had included some boilerplate SLA language that guaranteed certain levels of network and server availability but failed to guarantee what I was really concerned about – that my users were able to access the applications that are critical to our business with an appropriate level of performance, or response time. After all, what good is a network if you cannot access the applications running on it?

I am a firm believer that we need to raise the SLM bar by guaranteeing end-user quality of experience (QoE). For example, I am typing this article using Microsoft Word on a Citrix session. If the Citrix server crashes or is running too slowly for me to finish the article, I can’t do my job, and it wouldn’t make a bit of difference that the network is still up (other than the fact that I would be able to surf the Web instead). The SLAs proposed by my service provider would have done nothing to guarantee the performance and availability of my Citrix session, from _my_ perspective; they were written to guarantee availability only of a limited number of discrete components that make up the infrastructure.

Clearly, most service providers are motivated to provide SLM, but only to provide the bare-minimum service-level objectives that will make the sale to the consumer. It is up to the IT folks out there on the consumer side to demand more from their service providers. Once the providers realize they need to provide QoE-related service guarantees, they will take the steps necessary to actually provide it.

In the past, SLAs included “nuts and bolts” guarantees mainly because the technology was not available to measure real end-user QoE. Today, that is simply no longer the case. There are plenty of management products that monitor end-user sessions. One that I am familiar with is Reflectent’s EdgeSight, which does a great job of monitoring end-user applications and reporting problems back to a central console. Once collected, the metrics may then be used for SLM. This task should be even easier in a server-based computing environment (such as Citrix or Terminal Services), since the actual end-user experience is occurring on a monolithic server in a data center.

Unfortunately for me, our service provider was not able to provide QoE-based objectives because it lacked the necessary instrumentation (and we needed to get moving on our migration). But believe me, I am working very hard to help it raise the bar on its own SLM offerings.

I am very interested in hearing from vendors, service providers and consumers out there in the “real world” that are actively involved with end-user QoE-related service-level objectives and guarantees. I welcome your ideas, suggestions and comments on the subject of outsourcing at mailto:mehr@enterprisemanagement.com