Is there a future for Windtel? Part III (SUNW mentioned at end)
if anyone wants parts 1 & 2 i have archived articles....
September 27, 1999
Tawei Liao
This is the final paper in a three part series on Intel's new Internet Exchange Architecture (IXA) and WindRiver's VxWorks. In Part I, we gave an overview of the capabilities of IXA and VxWorks. In Part II, we talked about some of the problems that will arise Intel enters this market. This series will culminate with a discussion on the possible outcomes to the problems covered in Part II.
We recommend that you read Part I, Part II, and also our analysis of Intel's network strategy to get some background information.
Can the network equipment industry be "commoditized"?
Intel's IXP 1200 network processor combined with WindRiver's VxWorks real-time operating system is aimed to "commoditize" the network equipment industry. Intel executives have openly claimed that IXA will turn companies like Cisco Systems into the likes of Dell. But will that really happen? We believe that IXA will have a big impact in the small and mid-range segment of the market. But it will have little effect on the high-end segment of the industry.
The Internet has been growing at an exponential rate over the past few years. But the capacity of the networks has not been able to keep up. It is about time that Intel steps in and applies some hardware discipline to the industry. Moore's Law dictates that the speed of a microprocessor will double every eighteen months. Network processors will most likely follow Moore's Law as well. If Intel and WindRiver successfully deliver, and we think they will, the duo will be able to commoditize a big percentage of the industry. Any network equipment vendor that uses Intel's network processors is guaranteed a path of performance growth. Just like today's PC makers, they will be able to deliver a box that doubles in speed every eighteen months.
Using network processors will also reduce cost dramatically because it turns a hardware problem into a mostly software problem. Instead of piecing together many different hardware components to perform a task, designers can write software that performs the same task and run it on a network processor. It makes many hardware components unnecessary. The real-time operating system that runs on a network processor is essential. And in this respect, WindRiver could very well become the Microsoft of the network equipment industry. VxWorks could be to the network equipment industry what MS Windows is to the PC industry.
There is one segment of the market that Windtel will never be able to dominate: the high-end backbones of the Internet. This is the most important segment of the industry because high-end routers and switches command the highest gross and profit margins. Windtel's inability to commoditize this market is very important and warrants a close look. In order to understand how we came to this conclusion, let's look back at how the network equipment industry evolved.
In the early days of the Internet, there was no specialized equipment to handle Internet traffic. Routers were simply high-performance general purpose computers dedicated to running routing software. Sun workstations running the Unix operating system routed virtually all Internet traffic at one time. And then Cisco came along. The brilliant vision of Cisco's founders was seeing the need for a machine that just does routing. They stripped down a general purpose computer and wrote an operating system (later called IOS -- Internetworking Operating System) that does only one thing and does it very well: route Internet traffic. It became a huge hit. As Internet grew, so did demand for Cisco routers.
In time, technology improved and routers looked less like stripped down general purpose computers and more like specialized appliances. This next step in the evolution of network equipment came in the form of ASICs. ASIC stands for Application Specific Integrated Circuit. An ASIC is a hardware chip that performs the same task as software running on top of a microprocessor, or a network processor for that matter. The popular "Layer 3" LAN switches are powered by ASICs. These switches outperform routers because the routing protocols and algorithms controlled by software are moved down to the hardware. When silicon is "hard-wired" to perform only one task, it does so very very fast. Network processors will NEVER outperform ASICs because they are not "hard-wired" like ASICs. The flexibility of a programmable network processor makes it complex. That added complexity slows it down. Even if Intel produces a new generation of network processors that double in speed every eighteen months, it still would not be able to catch up to the speed of ASICs.
The high speed of ASICs comes at a very high price. It is extremely difficult to design and manufacture ASICs. If a bug is introduced in the process of turning a particular software functionality into a hardware component, the entire chip becomes unusable. Furthermore, ASICs cannot be reused. They are application specific and cannot be "ported" to another platform. ASICs work beautifully when designed and manufactured correctly. The ones inside the Juniper M40 router can handle 40 million packets per second. There is not a microprocessor or network processor in the world that can do that.
What does this all mean? It means that Windtel will do very well in the small and mid-range segment of the market. But it will never be able to commoditize the high-end network equipment market. Routers and switches designed with ASICs will always be faster than ones designed with network processors.
The Future of Cisco
The Windtel duo will put great pressure on Cisco. Vendors using IXA and VxWorks will undoubtedly start making "Cisco-compatible" routers and switches at rock bottom prices. How will Cisco compete with these clone makers? The future of Cisco will depend on their ability to tackle several problems. We will look at each of them one by one.
If you can't beat them, join them. In order to remain price competitive, Cisco must also start using network processors in their product line. And on the question of whether they will continue their relationship with IBM or switch to Intel, we believe that at some point they will switch to Intel. Not having "Inside Intel" is simply too risky. If they do not use Inside inside, they will miss out on the performance improvements that Intel is known to deliver.
Cisco will also have to make IOS into an open, modular, and fault-tolerant real-time operating system. This is an absolute requirement for the carrier market and also for a network processor-based product line. IOS as it is today is too big and too old to be ported to any network processor. But will Cisco rely on VxWorks since it is already in use in several of its products? We believe that VxWorks will still be used extensively in many of Cisco's products. But we can only speculate if VxWorks will serve as the foundation for an open, modular, and fault-tolerant IOS.
We have talked about the importance of ASICs in high-end systems. If Cisco wants to remain the 800-pound gorilla, they must make fundamental advances in ASIC design. That means developing advanced hardware technology in-house instead of buying startups. Cisco must be to ASICs what Intel is to microprocessors. For a more thorough discussion of Cisco's hardware strategy, please see our paper on Cisco's Strategic Inflection Point.
How will WindRiver compete with Microsoft?
When Intel network processors becomes the centerpiece of the network equipment industry, VxWorks will be right there with it. WindRiver will not have a problem competing against other RTOS vendors. They will have a problem with Microsoft. The size of Microsoft's booth in the exhibit halls of the annual Embedded Systems Conference in San Jose, CA is a clear indication of their intention. They want to dominate the embedded systems industry with Windows CE and Windows NT Embedded. The question then becomes how can WindRiver, or any company for that matter, compete with Microsoft? Anyone who has tried to compete with Microsoft has lost. Apple fought Microsoft in the GUI war and they lost. Netscape competed with Microsoft in the browser war and they lost also. The answer to this question is simple: don't compete with Microsoft, compete with yourself!
We all have much to learn from Intuit; they have been beating Microsoft for years. Intuit is the maker of Quicken and TurboTax. Despite numerous attempts to kill Quicken with Microsoft Money, Bill Gates still cannot get a handle on the personal finance software market. The success TurboTax has made the Redmond giant think twice about entering the tax software business. What is their secret?
They do not compete with Microsoft!
Your biggest competitor is yourself. You should be your harshest critic and you should only worry about surpassing your own very high expectations of yourself. A company cannot define itself through the eyes of its competitors. Such a company has no soul and no culture. It stands for nothing except for fear and hatred of its competitors. And eventually it withers away. Intuit did not become obsessed with beating Microsoft. It was obsessed with developing better products. Each version of Quicken was better than the previous one. Each version of TurboTax was better than the previous one. That approach worked; Quicken and TurboTax was and still is number one.
Apple and Netscape both succumbed to fear and hatred. They created entire industries but they ended up living in memories of past glories. Instead of relentlessly improving their products, they became complacent and let their guard down. Bill Gates once said, "success fools people into thinking that they cannot fail." Gates has become the wealthiest person in the history of mankind by repeatedly exploiting this weakness in his competitors. He does not always win of course, as in the case with Intuit. But he has beaten plenty of rivals who are consumed with arrogance, fear, and/or hatred.
Our answer might sound overly idealistic, but it is the truth. Let the competitive analysts in the marketing department worry about the competition. The rest of the company should be focused on one thing and one thing only: develop the best products in the world and compete only with yourself. It will take a lot of leadership from the Chief Executive Officer to make this happen. But if Intuit can beat Microsoft, so can WindRiver.
Will WindRiver be acquired?
If WindRiver Systems was for sale, who would be the most likely buyer? Some potential candidates are: Intel, Cisco, Microsoft, and Sun Microsystems. Let us look at each potential buyer one by one.
Intel is the obvious candidate. It chose to work with WindRiver to provide the first operating system to run on top of IXP 1200. It is not unusual for a hardware company to buy a software company to market its chips. For example, Texas Instrument spent $435 million acquiring Telogy Networks to promote the use of its DSP chips in the telecommunications industry. Telogy is a big provider of voice over IP embedded software that run on Texas Instrument's DSP processors [1].
One of the drawbacks about this potential merger is that Intel might smother WindRiver to death. They have absolutely no experience running a software company. Intel has a large software division, but it is focused mainly on microprocessor design tools. And this design software is for internal use only; they have never marketed or sold any software. If they buy WindRiver, that might just be the end of VxWorks.
Another problem is that buying WindRiver will hamper support for IXA in the rest of the RTOS community. Intel wants as much support for IXA as possible. But buying WindRiver will make Intel a competitor to all RTOS vendors. That might drive them to support IBM's network processors instead.
But all these problems might not stop Intel from trying. If VxWorks is to become the Windows of the networking equipment industry, Intel will most likely do something about it. Microsoft is the dominant partner in the Wintel relationship. They will not want to deal with WindRiver the same way they dealt with Microsoft. WindRiver has been without a CEO for almost five months. Only earlier this month did the board select a new CEO. That could be the only reason why Intel has not struck a deal with WindRiver.
Cisco is also a likely candidate. They can gain a lot by buying WindRiver. VxWorks could be the core platform used to modularize and open up IOS (see our Cisco paper for more information). But they have one big hurdle: The Federal Trade Commission. VxWorks is widely used among all network equipment manufacturers. Owning VxWorks would give Cisco too much control in the network equipment industry. The FTC has already probed Cisco once. They will not hesitate to stop any actions deemed monopolistic.
Microsoft of course can also buy its way into the embedded software market. But as in the case with Cisco, anti-trust issues come into play. With the current lawsuit pending, we do not think that Microsoft will do anything stupid.
The most likely candidate of the four is Sun Microsystems. Sun has been seeking to expand into the Internet consumer appliance market with its Java technology. What better way to put Java into Internet appliances than to buy "the leader in embedded development software and services for the post-PC era"? In addition to the consumer market, Sun has also targeted the telecommunications industry with a full line of products and services. Early this year, Sun introduced fault tolerant servers and a wide range of software for the telecom industry [2]. By acquiring WindRiver, Sun will be able to provide an end-to-end solution to network equipment manufacturers. Vendors will be able to purchase everything from fault tolerant servers to an embedded real-time operating system and a plethora of network related software.
Sun designs and sells microprocessors like Intel. But unlike Intel, Sun also develops software and sells end-systems. They know how to sell and market software. What Sun lacks is a decent real-time operating system and embedded software development expertise. They tried to expand into this market back in 1997 by acquiring Chorus Systems[3]. This acquisition was a spectacular failure. Sun moved in the right direction but they bought the wrong company. WindRiver is a much better partner for Sun. Their product lines complement each other since the VxWorks development environment runs mostly on the Unix operating system that powers Sun workstations. The two companies combined would create a lot of synergy.
We look forward to seeing if Scott McNealy, CEO of Sun, and Tom St. Dennis, CEO of WindRiver, share our vision of the future.
References
[1] Texas Instrument Press Release
[2] Sun Microsystems Press Release
[3] Sun Microsystems Press Release
Home | Feedback
|
