>> continued from previous page
So What's Policy?
To help clarify policy and related terms, the IETF has been working for more than two years on a document on terminology. RFC 3198 Terminology for Policy-Based Management parallels Merriam-Webster's definition of policy: "a definite course or method of action selected from among alternatives and in light of given conditions to guide and determine present and future decisions."
This definition is vague and does not help us develop much of an understanding of what we mean when we say policy management. There are, in fact, three different types of policy that people refer to when talking about policy management: policy-based configuration; configuration of policy; and finally, from the department of redundancy department, policy-based configuration of policy. To help beat the confusion, we explain each in depth in "IETF Wrangles Over Policy Definitions".
Policy is nothing new. We've been using a policy-based approach to management for years. What has changed is that we've begun to formalize how policy-based configuration can be done.
In the past, we had scripts and experts that "just knew" how to do it. Now, new protocols and new ways of looking at information have been created to accomplish this type of configuration. Products have begun to emerge using some of this new technology combined with traditional methods. Of course, some vendors have simply slapped a coat of paint on old approaches and called them new.
The essential distinction between how things have been done in the past and what PBNM products now offer is this: These new products (some of which we look at in "Orchestream Conducts PBNM With Precision,") aim to let local experts input their rules and policies into the software, which then executes these polices, saving script-writing time. This approach is also much more dynamic than configuration-file generation. In more advanced systems, vendors have begun to offer scheduling capabilities, which let different policies be loaded at preset times.
Class Distinctions
Vendors follow the money. If customers demand -- and are willing to pay for -- improved management software, vendors will provide it. The question is, What do you ask for? Or, more to the point, Be careful what you ask for, you may get it. If we press vendors for each new technology that comes out the door, equipment vendors will do their best to deliver but will have little time to develop these new technologies into full-grown management solutions.
There are basically two groups of policy vendors. The A List includes those that make a living building policy applications, while the B List comprises those that make a living selling hardware. A List vendors include Cisco Systems, Dorado Software, Gold Wire Technology, Orchestream, Syndesis and SmartPipe. The B List includes network infrastructure vendors, such as Alcatel, Extreme Networks and Nortel Networks. Cisco gets entrée to the A List because it's most often mentioned by the other A List vendors as being their main competition. The B List group is limited because its members handle only their own equipment, hardly strategic or realistic for a heterogeneous network environment. The exception to this is Extreme, which supports some Cisco gear in addition to its own equipment.
PBNM products differ from traditional network-management frameworks, such as Computer Associates International's Unicenter, Hewlett-Packard's OpenView and IBM's Tivoli, primarily because PBNM vendors are trying to solve the problem of network-infrastructure configuration, not the sweet spot for those old stalwarts, which is fault and performance management. A possible exception to this is Computer Associates, which has said it is considering a multivendor configuration product this year. If that happens, it will turn up the competitive heat.
Features that PBNM products do share with the framework daddies include network equipment inventories, polling engines for state monitoring, topology views, event handling, access control and proprietary APIs. On the other hand, PBNM products don't have wide third-party support, though their vendors are clearly looking to foster this integration by publishing command sets and JavaBean libraries and by making their products completely executable from the command line. Still, even if they start to offer heterogeneous configuration, it's highly unlikely that HP and CA will go head to head with PBNM vendors. It will always be about the central data store and single management view for the frameworks, while the PBNM crowd has to offer solutions that react more quickly to market changes by solving new problems.
It's in the Way That You Use Me
The most obvious use of PBNM products is within a service-provider network, where bandwidth is the product. Enterprises could benefit from having a way of quickly and assuredly getting their networks back to a known state or of turning on QoS for applications like VoIP (voice over IP), but the dollars are going to be softer in those cases. For service providers, which have real costs assigned to the complexity of provisioning ATM and frame relay circuits, these products have a definable return on investment. The price tags are in keeping with this goal of service providers, and policy-management vendors do view the service-provider market as their primary target.
Business is converging traffic types into a single network, which is being virtualized by way of VPN, MPLS and QoS technologies. These two forces are creating a management vacuum, which PBNM vendors are attempting to fill. The lack of standards is a serious hurdle, but for specific applications, real solutions are beginning to be provided by the vendors covered in our review.
RSS
