The Cisco ASR 1000 gives superstar software architect a chance to prove his radical concept for building the machines that run the InternetMarch 04, 2008
March 04, 2008
By Charles Waltner, News@Cisco
Michael Beesley is one proud father. And he should be. He just helped create what he believes to be the finest achievement of his career. "I've done this a long time, and I can say that by an order of magnitude the ASR 1000 is the best router I've ever created," he says.
Of course, Beesley is known to be a little outspoken on such matters. But as one of the most successful software architects in the networking industry, he can usually back up his opinions.
For Beesley, the Cisco Aggregation Services Router (ASR) 1000 Series is gratification with a capital "G." After struggling so long against doubters, he is clearly enjoying the moment. He has reached the summit of what he calls "a technical Mount Everest."
Certainly, other Cisco System executives share his enthusiasm. They say the ASR 1000 will not only inspire a new generation of networking equipment from Cisco, but it could influence how routers are designed throughout the industry.
Beesley says the ASR 1000 and its offspring have the potential to run most parts of a communications system, leaving only the very high and low-ends of a network to specialized equipment. Beesley considers Cisco's new flagship product so capable that he calls it the "Swiss Army knife" of routers. "It can do it all, and you can build a whole network out of them," he says.
"Edge" routers such as the ASR 1000 are on the networking hot seat nowadays. These devices do much of the detailed work for managing the extensive and rapidly growing array of communications options possible with Internet protocol (IP) networks. Edge routers, for example, are where networks fine-tune video and audio traffic to ensure high-quality images and sound. An edge router might also keep track of a wireless connection, making sure a software application doesn't crash even if the connection gets fuzzy.
These tasks, however, require routers to know much more about all the millions of digital packets blitzing by every second-what networking engineers call "stateful features". But all of this brainwork can dramatically slow down and even confuse traditional routers.
Enter the ASR.
One of the most noteworthy characteristics of the ASR 1000 is its ability to keep running no matter what, a capability known as "high availability." The router's software architecture, dubbed IOS XE, lets it run a second copy of the operating system as back up, and allows engineers to make modifications on the router without taking it off line. If the router does need to switch to the backup system, it can do so eight times faster than the blink of an eye.
The development team for the ASR 1000 also restructured the company's traditional Cisco router operating system, IOS, to take full advantage of the multi-core computing capability of the Cisco QuantumFlow Processor. Though full of new features, IOS XE still uses Cisco IOS for much of its basic functions and management interfaces.
Beesley is particularly proud of his router's unprecedented modularity. His group designed the ASR 1000 to be able to use the same "piece parts" for every model in the line. "Regardless of what role an ASR plays in the network, it will all look the same as its siblings and can swap parts so customers don't have to buy special components that can only work with one model," he says.
The new modular architecture for the ASR 1000's operating system also provides a flexible "pay-as-you-go" option, allowing networking managers the luxury of updating capabilities as needed over a decade or two. "You can continue to make it better and better over time without much heavy lifting at all," Beesley says.
While many routers now offer high availability, and other networking devices can run various stateful functions, Cisco says the ASR 1000 is the first to offer such effortless support for so many features combined with such bulletproof dependability.
AN ACADEMIC INTEREST
Given Beesley's career, his role with the Cisco ASR 1000 is not surprising. The Irish-born Beesley was one of the company's first engineers, and he led development of several of the networking giant's most successful router families, including the stalwart Cisco 7200 series.
In 1997 Beesley left Cisco to help start Juniper Networks, now one of Cisco's toughest rivals. There he played key roles in developing the company's Junos router operating system, as well as designing some of its highest-profile products, including the M20, M40, and M160 mid-range routers and the T640 core router. He estimates his work for both companies represents product sales in the tens of billions of dollars.
After leaving Juniper in 2004, Beesley started thinking about a new concept for a routerone that he felt would revolutionize the industry. It would be so flexible, intelligent and powerful that it could easily adapt to a vast array of networking roles while doing all of them far better than any other router.
I've done this a long time, and I can say that by an order of magnitude the ASR 1000 is the best router I've ever created.
To pursue his idea, he formed a small group of elite engineers and started his own company, BCN Systems, in April 2004. Beesley says he was driven in no small part by "academic interest" in seeing if such a router was even possible.
Being familiar with Beesley's engineering capabilities, Cisco became curious about his new project. Cisco, recognizing the potential of Beesley's idea, purchased BCN Systems in December 2004.
But even then, Beesley says, Cisco wasn't sure what to make of him. "Some executives called me insanely optimistic. Some thought I should be fired for promoting an idea that was going to fail. And at first even Mario Mazzola (the Cisco executive who approved the acquisition) said he wasn't quite certain if he should believe me."
But the one person who was completely confident in his concept was Beesley. "I never had a doubt," he says.
When Will Eatherton, the architect for the QuantumFlow Processor, first met Beesley, he, too, wasn't quite sure what to make of his boisterous colleague. "It was a kind of 'we're going conquer the world' view," Eatherton says. "But the more I've gotten to know Michael, the more he inspires me. It amazes me that across such a broad area he has such a deep understanding of how to go about build networking equipment."
Beesley says he will continue to oversee development of the ASR 1000 until it reaches "young adulthood." "I don't know after that. Maybe find the next big thing or maybe go to Disneyland."
Despite his egotistical bluster, Beesley apologetically says that the use of "I" is an Irish idiosyncrasy. "In Ireland, 'I' means 'we'," he says. "No one person can do this kind of thing. I might have had the original idea for the ASR, but it took hundreds of very smart people to make it happen."
SETTING A NEW FOUNDATION
Beesley says he got interested in computer technology as a child when his father bought a Texas Instruments calculator. Beesley couldn't put it down. At some point, he says, his parents "must not have been paying attention" and he somehow ended up skipping a couple of grades in school, eventually entering Trinity College in Dublin at age 15. By the time he was 20, he graduated with two bachelors' degrees and a master's degree in math.
Beesley says his first job out of college was perhaps the most influential for him and the one he credits with planting the seed for the ASR's software development methodology. He went to work at a small company in Japan run by a wealthy entrepreneur, who hired an elite team of engineers to build "super cool products" that may or may not have had any practical applications.
It was there where he realized that the best way to make computer-driven equipment was by writing the software first and then building the hardware to support it, an approach that contradicts the way most other computing devices are developed. "We've had the process completely upside down," Beesley says. "The software is the foundation and the hardware is the roof. People have been trying to build the roof first before there is anything to support it."
To develop the software for the ASR 1000, Beesley's team worked from hardware specifications and ran simulations to replicate how the software would function with the hardware. Besides providing superior results, Beesley says this approach is several fold faster than traditional hardware-first methodologies.
Realizing the scrutiny that will be placed upon this ambitious router, Beesley says he tested the ASR 1000 in every conceivable way, preferring to break it himself rather than have it show a glitch to a customer, something competitors would seize upon. "We knew we needed to be above reproach because the router's design is so radical," he says. "But no matter how much we slapped it around, it just sat there and smiled."
With his dream of the ASR 1000 finally realized, Beesley is smiling too.
Charles Waltner is a freelance writer in Piedmont, Calif.