Approaching the management of application-specific traffic through software rather than hardware can allow fine-tuning for optimum performance. Most network managers have, at some point, seen network suppliers' presentations depicting the history of networking. They start with the birth of the Ethernet and move through routing...
and switching, but where next? What is needed is an intelligent network, one that can make its own decisions about how to handle different traffic types on a dynamic basis. But is this fact or fiction? Well, a bit of both. There is nothing new about the basic idea of adding intelligence into the network. Token Ring had seven levels of prioritisation available in its specification, but no one has ever used them. And look what happened to that technology. The same fate befell Asynchronous Transfer Mode. With Ethernet there is much more of a backwash on which to compose than with either Token Ring or ATM - a simple starting point, in other words. Consequently, there have been a number of attempts - both IEEE standardised and proprietary, care of Cisco, 3Com and others - to provide some level of control over network traffic, both at the router/switch and at the client end of the flow, and with some degree of success. Layer 4 load balancers added a degree of intelligence to the process of directing traffic to the right server, allowing for one-offs such as traffic peaks along the way. However, basic traffic prioritisation and routing is one thing - adding real intelligence to the network is quite another. The problem is that the attempts at traffic management have been largely hardware-oriented, with little or no focus on the nature of the network traffic being managed, and certainly not application-specific. Current networking infrastructures will never resolve the kind of problems that have been around since networking began, namely that the network application traffic is, whatever the suppliers say, not network-aware, not network-friendly and not designed to run on networks. If it was, why would there be a huge market for enterprise software acceleration products? And here is where it gets interesting. Over the past couple of years a number of players have arrived on the Layer 7 traffic management scene, some concentrating almost exclusively on this very application, primarily through data compression. But what is really required is to be able to manipulate any application data at Layer 7 and talk directly to the applications and network services. Someone has to take responsibility for adding an injection of truly intelligent, flexible traffic management somewhere between client and host. Well someone finally has, with the latest wave of Layer 7 products. Having looked in our labs at products from F5 Networks in the US, and Zeus, a UK company - both completely new - we drew two simple but important conclusions. The first is that the required level of intelligence to control and network hardware and application software is now available. The second is that the future of networking is software, not hardware. That is not to say users do not need hardware, rather that it should be purely functional and support the brains of the network - the software. F5 and Zeus are both traffic management products based on software that can be used to get very fine control over individual traffic flows across the network. Zeus can be bought just as software via a free trial download from its website installed on a vanilla Linux or Unix server. So although it is a networking product, it is a far cry from the hardware-oriented world created by Cisco, Nortel, 3Com and such. In the router world the focus has switched between software-oriented and hardware-oriented products over the past 15 years, as the relative cost of Asics versus memory/disc space has itself varied, so a pattern has emerged to follow these trends. Here, however, the products are trying something far more fundamental: not a short-term trend based on economics and technological advancements, but a genuine revolution in the way networking traffic can be and is being managed - as fundamental a change as routing was when it was first introduced. It is not just a case of software being a superior solution, as it avoids the problem of limited lifespan, courtesy of the hardware and available bandwidth technologies. It means there are no physical limits to what you can do and where you can go within the limits of available bandwidth. Everything can be optimised, so every byte of bandwidth can be used. It gets even more interesting when networking applications are developed directly in conjunction with Layer 7 devices - a marriage of products such as Thingamy and F5. This kind of integration is only possible because the networking device is basically software. In this case, a combination of F5's iRules (essentially a programming language) and iControl (an open interface to the F5 devices supporting all the major development frameworks, such as .net and Java) means you can literally program networking. So a network flow can be tuned, based on the application or service conditions, helping to automate communications between third-party applications and the F5 system, eliminating the need for manual intervention. Similarly, with Zeus' ZXTM product, its Trafficscript rules language, which is very similar to Perl, enables you to interact with third-party applications directly, such as via a Soap/XML interface, and even call applications from within a rule, so that the Layer 7 device completely manages every application on the network. This is real network management. Here is another real application when networking gets intelligent and programmable. Forget the old marketing spins by play-safe suppliers, revolution, not evolution, is long overdue in networking, yet it need not be bloody. Here users have a situation where, with these kind of traffic management capabilities, it is possible to re-engineer the network completely, but over as long a period as you need. As such, this software programming approach to networking provides us with a true migration toolkit. With respect to IPv6, this enables moving between enterprise software applications, changing authentication servers and services, or any scenario where a change for the better should be made, but previously would have been too painful in terms of time and cost. As if to emphasise the change in direction of networking, F5 even has an online community - Devcentral - for its iControl and iRules developers, just as any programming language or framework would, for creating shared libraries of code, swapping ideas or problem sharing. This is a far cry from switch-back plane fibre and port density specifications. ATM failed to take over the networking world, not just because it was too complex, but also because it did not truly engineer a revolution in networking. Now, with Ethernet as a basic engine and networking as a programming language, real network applications can be developed from the ground up for the first time. Welcome to the future of networking.
Steve Broadhead, Broadband-Testing
Steve Broadhead runs Broadband-Testing Labs, a spin-off from independent test organisation the NSS Group. Author of DSL and Metro Ethernet reports, Broadhead is now involved in a number of projects in the broadband, mobile, network management and wireless Lan areas, from product testing to service design and implementation.