Network Management Strategies for the CIO
In Part two of this two-part series, we explored the difference between network and application awareness. Apart from the crude behaviors attributed to TCP or QoS, awareness is relatively
Requires Free Membership to View
new.
However, there are indications of a trend towards a unified "intelligent" network
infrastructure. In Part 2, the outward signs are examined and a little futurism applied to
figure out where this is heading.
Unification
The relationship laid out in Part two between network and application awareness is not accidental.
They represent two kinds of visibility -- one top-down (from the application) and the other
bottom-up (from the network). They are aimed at unification into a coherent approach to assured
application performance - in particular, for the benefit of VoIP. The drivers for this evolution
aren't limited to VoIP -- rather, VoIP has defined the necessary criteria for a successful
relationship between the network and application.
Given that application performance is a matter of quality of experience (QoE), defined either by end-user or by a business process, network mechanisms need to be more tightly bound to that experience. And regardless of how networks are implemented, some means of verifying their ability to support applications are needed. Inherent in both requirements is the development of models for application performance based upon IP behaviors.
New layer model
The end-to-end path is typically viewed at the level of IP -- or, in other words, it's all about
packets. How packets are behaving is a function of all the contributing factors at Layers 1, 2, and
3, from one end-host through all the mid-path devices to the other (see Figure 1 below). Media
(like wireless), NIC drivers, cross-traffic and a multitude of other forces (some dysfunctional)
affect the passage of packets. And directly affect the QoE of the application.
Layers 4 to 7 are typically localized at the end-host (whether IP phone, server, or workstation). Beyond the end-host lies the influence of other systems, work processes or the end-user -- the so-called Layer 8. For VoIP, Layer 8 is clearly the subjective psycho-acoustic influences of the typical human.
As shown in Figure 2 below, the "new layer model" for QoE can then be simplified to IP
behaviors, application/host, and the end-user or external environment. The goal in this exercise is
to ensure that networks effectively support a particular application such that it meets the
requirements of the end-user. Necessary to this framework is an effective (set of) application
model(s).
Application modeling
VoIP has done the network industry a huge favor (see VoIP:
The face of the new network police). It has defined an approach for linking network
behavior to application performance. A variety of means have been developed that perform that
function. And in the process, an effective means for implementing application models at the network
level has been developed. For example, network devices can assess MOS (mean opinion score) based on
the underlying packet behaviors -- MOS, of course, representing VoIP performance in terms of
subjective opinion.
Although MOS is an overly simplified metric (see MOS: A love-hate relationship) of limited value, it represents the core value of the underlying VoIP model. Breaking MOS into listening and conversational components (LQ and CQ) exposes more of that value. More importantly, this illustrates what is possible and that it is possible for other applications as well.
Intelligent networks
Given effective application models, the picture begins to resolve as shown in Figure 3 below.
Specific application models can define requisite IP behaviors -- as the recently released ITU
Y.1541 standard does -- for the purposes of QoS implementation and also for SLAs (see Death
of SLAs). And conversely, the measured network performance can be translated into an
effective QoE (MOS in the case of VoIP).
Today, the VoIP industry has developed effective means of doing the second step -- that is, MOS can be readily extracted from IP behaviors, whether by passively or actively sampling the network. Measuring MOS has changed from being unheard-of a few years ago to a relatively well understood feature of most VoIP infrastructures.
However, QoS mechanisms are still maturing (see Making the triple play). And the attention on network health has not borne the full extent of its fruits -- QoS won't work on a broken network -- so many networks are simply not ready for QoS. And reliable mechanisms that ensure end-to-end paths are not widely available for broadly heterogeneous networks. Finally, the prospects for "converged QoS" mechanisms -- QoS implemented for multiple applications with distinct requirements and metrics haven't begun to appear.
But once a robust system for coupling IP behaviors to application performance emerges, the foundation for a "smart," if not truly "intelligent," network will have finally been established. This coupling offers the visibility (today's "awareness" mechanisms) that is requisite for self-healing, autonomic networks that effectively support the applications that use them.
Autonomy in the sense of Step 4 of Gartner's IT maturity model will require much, much more: means for automated assessment, remediation and provisioning, standardized interfaces for querying network devices (SNMP must go away!), and SOA-style frameworks that integrate networks with application services. But that's a few years away.
For now, we will have to make do with "just aware."
References:
ITU Y.1541
NetConf
Chief Scientist for Apparent Networks, Loki Jorgenson, PhD, has been active in
computation, physics and mathematics, scientific visualization, and simulation for over 18 years.
Trained in computational physics at Queen's and McGill universities, he has published in areas as
diverse as philosophy, graphics, educational technologies, statistical mechanics, logic and number
theory. Also, he acts as Adjunct Professor of Mathematics at Simon Fraser University where he
co-founded the Center for Experimental and Constructive Mathematics (CECM). He has headed research
in numerous academic projects from high-performance computing to digital publishing, working
closely with private sector partners and government. At Apparent Networks Inc., Jorgenson leads
network research in high performance, wireless, VoIP and other application performance, typically
through practical collaboration with academic organizations and other thought leaders such as
BCnet, Texas A&M, CANARIE, and Internet2. www.apparentnetworks.com
This was first published in May 2006
Join the conversationComment
Share
Comments
Results
Contribute to the conversation