This article, with contributions from Dominic Orr, Selina Lo and Steve Wood, is posted with permission from Philip Goldie and Matthew Syme, authors of Optimizing Network Performance with Content Switching, ISBN 0-13-101468-4, published by Prentice Hall PTR, copyright 2004.
From its roots in the "Internet boom," Content Switching, or Layer 4 to 7 switching, is fast becoming a core component in the modern enterprise network. But how did we move from the world of the Internet data center (IDC) to the enterprise computer room? To help answer this question, we sat down with three of the people who were key in building content-switching pioneer Alteon WebSystems prior to its acquisition by Nortel Networks. Dominic Orr was the President and CEO, driving the company's eventual acquisition in late 2000. Selina Lo was VP of Product Marketing, leading the awareness of the technology in many of its early adopters. Steve Wood is currently President of Nortel Networks Australia, having previously led the Alteon business across Asia Pacific.
Two and a half years after the acquisitions Nortel and Cisco made in 2000, the landscape has changed considerably. The dot-com bubble has apparently burst and the financial dynamics of the carrier, hosting and ISP markets has shifted, presenting new challenges for operators and vendors alike. So how closely tied was the development and adoption of content switching to the original Internet Boom?
"There's no doubt that this explosion in Internet adoption and the rush of the bricks-and-mortar and dot-com companies to get online definitely helped accelerate the content switching market," says Dominic Orr, "but where the technology was used is key."
Selina Lo elaborates, "There are two parts to content switching: Firstly, traffic or sessions redirected based on pattern matching, and secondly, load balancing. Initially ISPs were using redirection to get around slow links and maximize expensive bandwidth. Three years ago, the hosting companies were reselling content switching services as part of a hosting package and were, by default, the distribution channels for the enterprise. We have now seen a shift emerging where enterprises are managing some of this process themselves."
As some of these enterprises have looked to move out of the IDC and into their own data centers, the industry has seen more adoption of the technology into the core of the enterprise networks, which in turn is driving more and more of its use. "Since the dot-com era, the focus of content switching has shifted and is now the cornerstone in opening up the data center to the enterprise world," observes Orr.
Lo agrees: "That content switching is a core technology is a given. This can be seen by the uptake and adoption by forward thinking enterprises. This is giving them a competitive advantage in their day-to-day business and how their users interact with the applications and data that bring in the revenue. You know, content switching is not a poor man's high availability (HA) as there are other advantages as well -- it gives enterprises 90% and more of the HA functionality they're looking for, but for only 20% of the effort."
But what are the some of the next stages of development required to further drive the technology deep into the enterprise network? Well, greater integration with the next generation of enterprise applications is clearly one such step. "The big difference between L4 and L7 is that above L4 each application has its own standard in terms of headers, ports, etc.," says Lo. "Content switching is very application-dependent, and you have to customise the content switch to the application. Standard applications such as FTP and HTTP are easy to do because their standards are clearly defined, but more complex applications such as Citrix and Oracle require a lot more configuration. Application vendors will be interested in working with the content switching vendors as more and more customers force them to do so," says Lo.
"Absolutely," agrees Orr. "Application vendors have traditionally paired with server vendors rather than network vendors, and as server vendors integrate content switching into their hardware (such as blade servers or chassis-based servers), this will force application vendors to follow suit."
Back in 2000, Cisco Systems and Nortel Networks raised the stakes and aggressively entered the content switching market with their acquisitions of Arrowpoint Communications and Alteon WebSystems, respectively. Now, three years later, do those acquisitions seem to have had an impact on validating or driving the technology?
"In terms of validating the technology, this had largely been done by many of the early adopters. One thing it definitely did is help bring the technology to the masses and certainly in that category were the medium and large Enterprises" says Steve Wood. Certainly the early adopters had helped take some heat out of the technologies infancy agrees Orr. "As the technology and focus shifted from the ISP to the enterprise the acquisition allowed the technology to be more readily adopted by the more conservative enterprise market as it was now branded and backed by the industry leaders in enterprise networking". Going forward, the integration these and other companies are doing is helping also. "If you look at the work Nortel, Cisco and others have done in integrating Layer 4 to 7 services into their existing Layer 2 and 4 platforms, it again really points to this whole concept of validating the technology as something core to modern networks" say Lo. So, given some of the challenges of where the technology fits, how difficult did it prove for start-ups like Alteon to recruit the new breed of Sales, Marketing and Engineering staff to drive the company? Indeed, this is an issue that many enterprises are grappling with even now. "A major challenge was how to bring the server to the network and the network to the server," Orr says. "Most server people were uneasy about this technology being on the network equipment and did not want this. The struggle was to get people who understood this technology, as most understood either TCP/IP or the applications that the servers ran, with very little overlap. It was those engineers who understood routing and Unix based systems that were able to grasp this technology. This was a breed of person for which we coined the phase 'Web Heads.' They were able to internally cross breed and train others with their skills and get the server to interact with the network."
Selina Lo agrees that this breed of network and application-aware Web Head is someone hard to find even in today's market. "The key to this was education," she says. "It is still a new and fairly complex technology, and in the early days we needed to mobilize the Web Heads to sell this technology up to the management and application people. It is difficult to find Web Heads, as they don't fit into any specific place within an organization because their skills transcend many disciplines. However, all enterprises need infrastructure, DNS and firewall experts."
Orr suggests that enterprises should be looking to take this one step further and develop the skills of people able to speak the language of both the network and the applications that run across it. "Enterprises should identify these Web Heads and encourage and nurture their skills as they will need specialists who can transcend the technology and know how to construct an infrastructure from hardware to application-based services," he says. "It is similar to the migration of skills 10 years ago, when engineers moved from SNA to TCP/IP."
Where does the discipline of content switching find itself in 2003? Our panel was in no doubt about just how important the technology will be moving forward.
"Content switching is the new infrastructure. Traditional Layer 2 and Layer 3 switching vendors must expose their equipment to cater to applications and services. Like it or not, people will have to accept that this is the way forward," summarises Lo.
Likewise, Orr offers no hesitation in defining its importance: "A content switch is the body of a switch with the soul of a server. It is the very glue that will tie together the next-generation best of breed networking from servers to databases, wireless to storage networks. Organizations with the glue will succeed, while those without will be left behind."