IT'S IT NOT IS

While Information Services (IS) exists as a term, there is a reason that the infrastructure, applications and protocols that make up modern enterprise networks are collectively called Information Technology (IT). High level manager-types today try to talk in terms of offering services to end users. They try to define service models with Service Level Agreements (SLA) that provide end users with some sort of satisfaction that things will work as they expect.

Information Technology is made up of technologies – not services, so to try to characterize the technologies as services is not only inappropriate but also dangerous. Nowhere is this more evident than at the most basic level that glues the enterprise together – the network.

The network is the most abstracted layer when IT is talked about in terms of services. A service – say global videoconferencing –relies on a set of processes, procedures and Service Level Agreements to be offered to end user groups. This abstraction level relies upon a vendor and their physical assets and protocols – such as Polycom using the VSX product line with H.323. Finally, this "solution" must run on the network – the most basic building block. The network is regarded as electricity – plug in the computer and it comes on like a light. While this can be expected of a well designed infrastructure, the fact is most enterprises have a cobbled framework of interconnected devices that supports minimum IP connectivity between remote sites.

I believe this is the result of the paradigm shift away from the "T" (technology) and into the imaginary "S" (services) that exist nowhere in the name of the job description that managers in the field fulfill. Instead of improving technologies that will serve to create the foundation for the elusive services yet to come, they strive to exceed user expectations by promising wildly optimistic visions while also committing to cost cutting and staff reductions in an effort to appease management. The drastically overworked and underfunded IT professionals are soon pushed beyond their breaking points and the inevitable discussion of outsourcing rears its ugly head.

But is the problem the IT worker? Is it the IT managers? Is it the high-level muckity-mucks that demand more for less? The outsourcing model would have you believe the problem is the network. After all, that’s what is always outsourced – along with the jobs of those who run it.

Assume the "network" is the problem - after all, I've already stipulated that most enterprise networks aren't the case studies for proper vendor-endorsed showcase examples. But the network works - again, we've stipulated basic end-to-end IP connectivity exists. And any network architect worth his or her salt can make a poorly performing network perform better given the right amount of resources – people, talent, time and of course – money.

Therefore, don't outsource the network, outsource the problem. Again, the problem is not that the network can’t deliver PINGs from end to end. It's that the network can't deliver a high-quality, consistent and reliable experience for a particular application for user A, B and C. That isn't a "network" problem. It's a quality problem. And in network terms, quality is addressed with Quality of Service (QoS).

QoS – in some circles – is a show stopper. This is mainly because those talking in terms of "services" don't understand how "technology" makes their services possible. It's like trying to explain an engine tune-up to a beauty queen.

QoS isn't (just) queuing, marking and Differentiated Service Code Point (DSCP). These are technical components of a QoS "service" (See, we can talk in terms of services – when appropriate). QoS is an umbrella term that incorporates a suite of technical components to address all concerns in network transport. Congestion management and avoidance, call admission control, priority queuing for real-time applications and link efficiency mechanisms for low-speed Wide Area Network (WAN) pipes are all pieces of the QoS puzzle. Given the features and depth of a QoS solution, you don't turn it on and walk away. You need planning and global consistency. You don't implement everything; you tailor the solution to fit the requirements – and just so you management-types don't get confused, we're NOT talking about business requirements, we're talking about technical requirements.

Link efficiency mechanisms make T1's look like T3's. Queuing allows Voice over IP to perform like a normal landline call during both normal and intense network congestion. Advanced buffering algorithms detect network congestion before it happens to minimize impact. Advanced call admission control techniques ensure that limited bandwidth is not oversubscribed by bandwidth intensive applications. The end result is a tuned network that performs like a dragster rather than a drag.

If management-types need a "service" term to explain this to even higher management-types, talk in terms of Quality of Experience. QoE can address the non-measurable but required metrics of end user satisfaction in terms of consistent reliable performance for a given application or service. QoS is the tool of the "technology" to achieve the performance of the "service".