One World, One Processor?
One of the great advantages of Linux is that it runs on practically any processor you can find. Most competing flavors of UNIX are limited to a single CPU architecture. But in the server world, processor choices are diminishing. While this may not hurt Linux, it certainly doesn't help. These changes have been driven by Intel, which would love to turn the server market into a clone of the PC market, with hundreds of vendors selling similar systems based on Intel silicon.
Although far more PCs than servers are sold each year, the revenue from these two markets is actually similar, since servers are much more expensive, and have higher profit margins. So if Intel can grab its typical share of those profits (that is, most of them), it would be a great coup for the company.
So far, this strategy is off to a great start. In 1998, Intel launched a one-two punch, introducing its first Xeon processors designed exclusively for servers and disclosing the design of its Itanium processor for very large servers. With the exception of Sun, all major server vendors quickly adopted a two-track strategy, offering customers the choice of either Intel-based systems or systems using in-house RISC processors. The problem is that customers, for the most part, have been choosing the Intel-based systems, which use standard operating systems (Windows 2000 or Linux) and offer compatibility with systems from a variety of vendors. The RISC systems tend to be more expensive and rely on proprietary UNIX operating systems. Intel now holds nearly all of the market for low-cost servers (under $10,000) and more than half of the market for more expensive systems.
This market shift has left the RISC vendors with less revenue to support their processor lines. During the same period, high-end microprocessor designs have become more complex, requiring a greater investment to develop new products. With these two trends moving in opposite directions, Compaq, HP and SGI have slowed the pace of their RISC architectures (Alpha, PA-RISC and MIPS, respectively) to the point that they have fallen behind Xeon in performance on many key applications. This decline has caused a downward spiral in sale for the RISC vendors.
As a result, HP and SGI have already announced they will eventually discontinue their RISC lines, and Compaq is likely to follow suit. IBM continues to invest in its PowerPC line, but the bulk of the company's servers rely on Intel processors. At the recent Microprocessor Forum, the CPU industry's premiere event, IBM was the only one of these four companies to present a paper on a future RISC processor.
Sun has avoided this slippery slope by focusing exclusively on its SPARC/Solaris platform, and the company has never been healthier. But even with about 20% market share for servers above $10,000, the company is having problems developing new processors. It recently rolled out its UltraSparc-3 processor nearly two years behind schedule and announced an 18- to 24-month delay in its plans for UltraSparc-4 and UltraSparc-5. Intel solves this problem by leveraging the same processor designs between its PC line and its server line. Thus, it can invest far more than any other company in the design of its processors. As CPU designs become more complex, this level of investment is needed to maintain competitiveness.
Even as Intel dominates the processor market, it splits the market between two different architectures: Xeon and Itanium. But so far, Itanium has been a dud. After many delays, the new processor should finally appear in systems over the next few months. But sources indicate its performance is disappointing, and most vendors are sticking with Xeon. Intel's new Pentium IV technology will appear in the Xeon line in early 2001, giving that line a further boost. Meanwhile, Itanium's hopes are pinned on a next-generation version that isn't likely to appear until 2002. Until then, Itanium will be a niche player.
AMD is gearing up to push its Athlon processor into the server market. If AMD succeeds, this move will put pressure on Intel to cut Xeon's price and boost its performance, much as the company was forced to react to Athlon in the PC market this year. Since Athlon and Xeon both use the same x86 instruction set, AMD's entry will strengthen that platform and make it even more successful than it is today.
Thus, for the next few years, Xeon and compatible processors will power the vast majority of servers. This convergence benefits Microsoft, which is focused exclusively on that platform, and takes away one of Linux's selling points. Linux must go head-to-head with Microsoft in features and performance to continue its gains in the server market.
|Speed Up Your Web Site with Varnish||Jun 19, 2013|
|Non-Linux FOSS: libnotify, OS X Style||Jun 18, 2013|
|Containers—Not Virtual Machines—Are the Future Cloud||Jun 17, 2013|
|Lock-Free Multi-Producer Multi-Consumer Queue on Ring Buffer||Jun 12, 2013|
|Weechat, Irssi's Little Brother||Jun 11, 2013|
|One Tail Just Isn't Enough||Jun 07, 2013|
- Containers—Not Virtual Machines—Are the Future Cloud
- Non-Linux FOSS: libnotify, OS X Style
- Linux Systems Administrator
- Lock-Free Multi-Producer Multi-Consumer Queue on Ring Buffer
- Validate an E-Mail Address with PHP, the Right Way
- Technical Support Rep
- Senior Perl Developer
- UX Designer
- Introduction to MapReduce with Hadoop on Linux
Free Webinar: Hadoop
How to Build an Optimal Hadoop Cluster to Store and Maintain Unlimited Amounts of Data Using Microservers
Realizing the promise of Apache® Hadoop® requires the effective deployment of compute, memory, storage and networking to achieve optimal results. With its flexibility and multitude of options, it is easy to over or under provision the server infrastructure, resulting in poor performance and high TCO. Join us for an in depth, technical discussion with industry experts from leading Hadoop and server companies who will provide insights into the key considerations for designing and deploying an optimal Hadoop cluster.
Some of key questions to be discussed are:
- What is the “typical” Hadoop cluster and what should be installed on the different machine types?
- Why should you consider the typical workload patterns when making your hardware decisions?
- Are all microservers created equal for Hadoop deployments?
- How do I plan for expansion if I require more compute, memory, storage or networking?