Policy-based networks: Easier said than done

If you listen to vendors and analysts, the Next Big Thing is convergence, that magical, mystical merging of voice, data and video onto one high-speed IP-based net.

In theory, it sounds great. You reduce the complexity of your network, you save money by running voice over IP and you get to introduce spiffy new video applications.

But the devil is in the details, and if you take a hard look at what's entailed in implementing IP-based quality of service (QoS) through policy- based networking, your head may start to spin like Linda Blair's in "The Exorcist."

For starters, a typical policy-based network includes a policy console, a Lightweight Directory Access Protocol-based policy repository, a policy server and a policy enforcement point (PEP). The policy server talks to the PEP through a new protocol called Common Open Policy Service. And, of course, you need a policy server, policy repository and PEP at each remote office.

Once you've got all that set up, you need policies. Policies are business objects translated into sets of rules that are represented as data structures, following the Policy Framework Core Information Model. Hey, no problem.

And once you've got all that sorted out, you have to implement draft QoS standards such as Differentiated Services (Diff-Serv), Resource Reservation Protocol and 802.1p. Under the Diff-Serv model, it's up to the network manager to configure algorithms on an interface-to-interface basis so all packets are classified, marked, policed and shaped. That's where you get to choose between algorithms such as weighted fair queuing, class-based queuing or the ever-popular class-based weighted fair queuing.

How many network managers will it take to set this up? How many will it take to monitor these systems, create access control lists, set algorithms and policies, and change policies?

And that's just the technical part of it. Let's talk politics. Who decides whether an SAP user is more important than an IP phone user? Which traffic gets higher priority, video of the monthly sales meeting or the CEO's e-mail? And who breaks the news to the CEO?

Convergence may eventually provide real business benefits, but at this point I wouldn't bet my network on immature products, untested protocols and evolving standards. Especially in a marketplace in which prices for traditional voice traffic and WAN bandwidth are falling.

- Neal Weinberg

Features editor

nweinberg@nww.com