Issue No.11 - November (2007 vol.40)
Published by the IEEE Computer Society
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/MC.2007.396
Topics covered include a desktop high-performance computer, a controversial wireless-spectrum auction, new technology that creates e-wallpaper, market growth in the thin-client concept.
Professor Builds Desktop High-Performance Computer
A US professor has built a high-performance computer that sits on a desktop and cost only $2,470 to construct. It now could be built for just $1,255.
Calvin College professor Joel Adams and former student Tim Brom began constructing Microwulf late last year. Microwulf, the name of which derives from NASA's 1994 supercomputer cluster known as Beowulf, performs at 26.25 gigaflops.
Microwulf measures just 11 × 12 × 17 inches. The increasing power and decreasing size of the parallel-processing multicore chips that Microwulf uses enabled Adams to make the computer fast yet small.
Adams said that Microwulf might not be the world's smallest or least-expensive high-performance computer but that it does offer the best price-performance ratio—$47.80 per Gflops based on current construction costs. He compared this to Stanford University's Bio-X supercomputer cluster, for which the price-performance ratio is $109 per Gflops, considered very efficient.
"The same way the PC made it possible for people to not have to wait on a mainframe [to perform computing tasks], the Microwulf design has the potential to let each person have a cluster on their desktop," he said.
According to Adams, two technology improvements made the project possible: multicore CPUs and Gigabit Ethernet.
Multicore chips, which have multiple computing cores on a single piece of silicon, improve performance by breaking tasks into pieces and performing them in parallel. Microwulf has four dual-core chips, which communicate via Gigabit Ethernet.
Adams has ported 12 applications and his students have written two demonstration programs to show Microwulf's capabilities.
In addition, Calvin College plans to use the computer to teach students how to write applications for parallel processing, which is becoming increasingly important as multicore chips become more popular.
The computer is also powerful enough to run complex computations for problems such as molecular modeling and computational fluid dynamics, according to Adams.
Future applications for Microwulf will probably focus on complex engineering- and science-related problems, which could be solved using the many free, open source, parallel programs that already exist for Beowulf clusters, he noted.
The computer has attracted attention from around the world. For example, Adams noted, a Volkswagen Group subsidiary in Spain, Centro Técnico de SEAT, asked for more information, saying Microwulf could let engineers have their own desktop high-performance machines rather than having to wait to submit designs or drawings to a central computer.
According to Adams, he built Microwulf to kill time between projects building cluster computers under a US National Science Foundation grant.
Important Wireless-Spectrum Auction Generates Controversy
An upcoming US Federal Communications Commission auction of radio spectrum highly valued by wireless-service providers includes a controversial new provision.
Starting 24 January 2008, the FCC will auction five blocks of spectrum, totaling 62 MHz in width, in the 700-MHz frequency band. This could be used by mobile telephones, which currently don't operate in this frequency in the US. United States UHF TV stations will abandon this spectrum when they move from analog to all-digital broadcasts by February 2009, as federal law requires.
Different sets of spectrum blocks will be available in different regions within the US, noted Allen Nogee, an analyst with market-research firm In-Stat. Thus, he explained, a service provider could win a specific block in one region but not in others.
The available spectrum will include a 22-MHz-wide block available in six regions. It will be in particular demand because it is the widest of the blocks, noted Nogee. Wide blocks increase the available bandwidth, and thus also let carriers serve more customers, by transmitting more data at any one time than narrow blocks.
Unlike the other frequency segments up for auction, this block is being reserved by the FCC for open access, the first time this has been done in the US. Service providers couldn't restrict the types of devices that use the spectrum, as long as the FCC has certified them for use there. Wireless providers can currently limit the handsets and applications that their customers use to those the carriers provide, rather than letting consumers choose them freely.
Said FCC chair Kevin J. Martin, "A network that is more open to devices and applications can help foster innovation on the edges of the network. And it will give consumers greater freedom to use the wireless devices and applications of their choice."
Carriers such as Verizon Wireless have opposed this proposal.
Opponents say the government should not impose these types of operating conditions on carriers. They contend this would hurt providers' ability to optimize the hardware, software, and network experiences of customers and limit the introduction of new and innovative wireless services.
"We are concerned that a significant portion of this valuable spectrum will be encumbered with mandates that could significantly reduce the number of interested bidders," said Steve Largent, president and CEO of CTIA-The Wireless Association. "We remain committed to the principle that wireless consumers and American taxpayers are best served when such a valuable commodity is auctioned … with no strings attached."
Companies such as Frontline Wireless and Google support open access.
Frontline chief technical officer Stagg Newman said, "Wireless broadband is the next frontier of innovation in devices and applications in communications and computing. Today, though, the US is falling further behind Asia and Europe in wireless innovation. Open access, one of the foundation principles of the wired Internet, is the critical ingredient that has enabled [competition and] innovation and is needed for some of the new wireless broadband spectrum."
The auction's minimum acceptable bid for the 22-MHz wide blocks will be $4.6 billion. However, the FCC has said that if no one offers at least the minimum bid, it will auction the blocks again without the open-access requirement.
The different spectrum blocks to be auctioned will be best-suited for different wireless technologies, based on the bandwidth they offer and other technical characteristics.
Will Thin Finally Be In?
A decade ago, companies such as Oracle and Sun Microsystems pushed the thin client—a stripped-down machine that uses network-based, rather than host-based, programs and storage resources—as a more efficient way to perform enterprise computing. Proponents also saw it as a way to break Microsoft's Windows-based hold on personal computing.
The thin-client concept never exploded, although, noted Bob O'Donnell, vice president of market research firm IDC's clients and displays program, sales have grown steadily for a number of years.
Now, though, market growth may be about to accelerate, according to Jeff McNaught, chief marketing officer at thin-client vendor Wyse.
IDC predicts 3.5 million thin-clients will be shipped in 2007, 6.8 million in 2010, and 8.5 million in 2011.
Over the next decade, thin computers could replace as many as 30 percent of all business PCs, said Tad Bodeman, Hewlett-Packard's global director of thin-client product marketing. Strong markets could include healthcare, financial services, education, and retail.
In light of this, HP recently acquired thin-client maker Neoware for $214 million.
Progress in hardware, software, and networking have made the use of thin clients, also called network computers, more like that of PCs for many tasks. For example, thin clients—which can be desktop- or laptop-based—now have enough processing power to render complex images quickly. And today's faster networks can transmit large files or make complex applications available to the machines.
Proponents say thin computing's key advantages include lower cost of ownership and lower energy consumption. Thin clients don't have either powerful processors for local computing or many of the components, such as hard drives, found in PCs. They are thus less expensive to buy, manage, run, and maintain.
Security is also a key issue, noted O'Donnell. Many employees now use laptops, which, if stolen or lost, yield confidential information that users sometimes store on hard drives. This wouldn't be a problem with thin-client laptops, which don't have hard drives.
Also, because network computers don't store programs that are hackers' targets, they are less susceptible to malware and spyware than PCs.
Proponents are no longer marketing network computers as a way to fight Microsoft. Instead, they are pushing their use in niches where they might offer advantages.
Bodeman said thin clients could be used as point-of-sale devices in retail locations, kiosks in public centers providing information about government services, and even communications devices in military deployments. These uses are well-suited for simple, low-maintenance machines without powerful local-processing capabilities.