By Michael J. Martin
Today, application-aware traffic management technology falls far short of what it must really offer. What we need is session-based application-aware routing – or the ability to granularly control multiple applications on multiple paths simultaneously. But that is only partially available from vendors today. Now more than ever it is time for real application-aware networking to emerge.
What is session-based application-aware routing?
Session-based application-aware routing would enable you to maintain symmetric application session flows operating within a bandwidth or service-level contract in a multi-path network with the ability to move application traffic between paths when necessary. That way, you could have multiple network links, each with its own operating heuristics, but you could manage application traffic so that overflow from one link could be moved to another link without affecting sessions or users on the first link.
Are your vendors dodging your questions?
So what's missing from application-aware traffic management?
It's currently possible to enforce the IP path that application traffic takes using policy-based routing on an edge or core router, or by utilizing a bandwidth-driven application-aware traffic management solution. But these two approaches fall short of the trinity: cognizance of session-state, application-traffic behavior and utilized path bandwidth simultaneously. The former solution is really an all-or-nothing sledgehammer approach, while the latter introduces asymmetric routing conditions in high-utilization environments.
It's odd that the overwhelming prevalence of Full-Duplex LAN transport has failed to spark any major realization that LAN and WAN networking today is becoming less about bandwidth and more about the optimization and symmetry of the uni-directional paths.
With latency-sensitive commodities trading, high-performance computing, and the World Wide Web of porn, asymmetric routing leads to roundtrip latency imbalances and increases the opportunity for jitter. Both conditions are deadly to any real-time application such as voice or video.
What's more, with inexpensive Gig and 10 GbE WAN transport readily available and capable of connecting locations separated by long geographical distances, a network admin's awareness of the real and discrete nature of these high-speed, high-capacity transmission paths is vital to delivering a reliable network solution. The reality is that a bottleneck on the client-to-server path can be just as performance-affecting as one from server to client. Lost TCP acknowledgments are just as hard to recover from as lost TCP data. Hence the need for application session awareness coupled with a high-performance policy routing engine.
But that's not all. As converged services tied to a single source/destination IP address continue to expand, network admins will continue to struggle with real limitations in their ability to maximize the bandwidth they have while providing quality real-time voice and video services. Call Admission Control (CAC) is largely an application-centric endeavor, if it exists at all. Protocols such as RSVP and RTCP exist, but there seems to be little desire on the part of the vendors to implement such functionality in their current products. The lack of a multi-application – dare I say "open" -- CAC solution, along with the limitations that exist in the current breed of server and host operating systems to effectively manage real-time application demands, only further illustrates the need for an application-aware networking solution.
About the author:
Michael J. Martin has been working in the information technology field as a network and Unix system admin for nearly 20 years. His early experiences designing, implementing, and supporting MIS infrastructures for research and ISPs give him a unique perspective on his current pursuits in large-scale internetworking and security architecture. As a network architect, he has designed high-speed/high-availability LAN/MAN networks for companies such as ANS/AOL, Philips, and the Edgix Corporation, and has provided network consulting for a number of businesses and regional ISPs. Michael also writes and provides training on networking and security related issues. His book Understanding The Network: A Practical Guide to Computer Networking has been translated into Russian and Chinese.
- End-to-End Network Application Performance Monitoring –Cisco Systems, Inc.
- Network Telemetry and Monitoring Agent for Application Performance –Apcela
- Software-Defined Networking for Better Application Performance in Healthcare –Comcast Business
- The True Cost of Application Performance Degradation –Accedian