The Race To Dominate High-Layer Processing
While the initial shots in the war to own the L4-7 processing
market are fired with silicon and hardware, it will be won with software,
tools and ecosystems
by Lee H. Goldberg
Back in 2003 when I awarded Seaway Networks one of analogZONE's Product of the Year Awards for its SW5000 Network Content Processor (NCP) they optimistically predicted hundreds of applications that would require deep packet inspection, content-related routing decisions, and something they called "application aware packet processing". But while I thought their novel architecture and unusual capabilities deserved recognition, I assumed that it would be lucky to end up in a handful of niche products. Back then I could only imagine a relatively small market for the high-performance, high-capacity media servers or high-security gateway and firewalls where the chip's high price and equally steep power requirements were just the cost of doing business.
It turns out that we were both right.
For the first few years, Seaway enjoyed only modest sales in highly-specialized applications (many of which I'm not allowed to talk about). But lately it's looking as if their vision of the future was simply a bit far-sighted as L4-7 processing is becoming a hot topic in mainstream networking. And it took me the same two years and two recent developments to finally appreciate Seaway's point of view myself.
It turns out that Seaway's concept of the widespread use of L4-7 processing has quietly eased its way into the mainstream over the past year or two as packet classifiers and traffic managers began to peer more deeply into packets, reaching beyond basic header elements to gain insights on how to treat time-critical, and mission-critical multimedia data streams. More recently Cavium Networks further validated the concept of content- and application-aware processing with the release of its Octeon processor series (reviewed here this week) which can perform wire speed L4-7 inspection, tagging, and even modification of the contents of multi-Gigabit multimedia streams.
This, and Freescale's announcement of its acquisition of Seaway this week, are clear signals that high-layer, content- and application-aware processing have entered the mainstream market. Even more telling is the fact that rather than adapt Seaway's content processing technology for their high-end C5/C6 network processor family, Freescale's stated intention is to integrate it as an on-chip engine (or series of engines) into their lower-cost, high-volume PowerQUICC series. See the tail end of this week's review of the Cavium chip for more of the juicy gossip about Freescale's plans for dominating the L4-7 processing arena.
Well, it turns out that the kinds of L4-7 processing that I scoffed at as unnecessary only a few years ago are rapidly becoming essential to support differentiated multimedia services that allow carriers to offer tiered service levels with varying levels of capacity, quality, and priority. The same ability to look deeply into packets and understand streams at the application layer are also key to staying one step ahead of increasingly cunning hackers and their sophisticated arsenal of viruses, spyware, and intrusion techniques.
I expect that over the next year or so, we'll see most of the other networking silicon makers trotting out their version of high-layer capabilities. And much like the encryption functions that proceeded them to the market, high-layer processing will probably follow the traditional migration path where it starts life running in software on general-purpose processors, moves to expensive standalone specialized chips, and finally ends up in the mainstream as compact, specialized accelerator cores running inside a larger standard processor or other commodity networking product. Depending on how they integrate high-layer processing into their product lines this new must-have capability could either help current market leaders secure their top-dog position, or give insurgents the opportunity to re-open markets by offering better solutions to these new challenges.
But silicon will not be not enough to win the race -- in fact it's only half of the equation. As we've seen from past waves of technology, one of the keys to winning market share is leveraging existing development resources as much as possible. The companies that do the best job at making high-layer content and application processing look like any other application within their development suite, and provide hooks to as many open source tools as possible will likely be the ones at the top of the heap at the end of the day. Silicon makers should take note of the fact that while innovative architectures, blinding performance, and breakthrough technologies always earn my admiration and respect, it's often the company with the most diverse and open development ecosystem that wins the hearts and minds of the folks who buy the silicon these days.
Comments? Questions? Exclusive tips on the first Layer-8 processor? Write me at: lgoldberg@green-electronics.com