Dr. Dobb's Journal February 2000
Enterprise Resource Planning (ERP) software has become the trend du jour in the business world. Tired of disconnected, outdated, and often incompatible applications performing various business functions, organizations have been looking for "integrated solutions," which do everything from inventory to distribution to human resources and accounts payable. Big players in the ERP software and services market include the likes of SAP AG, Baan, Oracle, and PeopleSoft, among others.
But many companies are finding out the hard way that massive and expensive application integration is easier said than done. In the past few months, several large, visible companies have run into serious problems with their ERP software. Hershey Foods' $112 million R/3 software from SAP AG fouled up the company's candy shipments for Halloween. Whirlpool's distribution system for its appliances suffered similar problems resulting from an SAP R/3 installation. Allied Waste Industries cancelled a $130 million SAP R/3 project, while Waste Management did the same after spending $45 million of an expected $250 million on a similar project.
Similar problems struck PeopleSoft, which has been sued by W.L. Gore (makers of Gore-Tex) for an allegedly botched ERP project. And Procter & Gamble's internally developed SourceOne global ERP system has been spitting out incorrect data and suffering from slow data retrieval times. There's no moral to the story. But caveat emptor when the big suits come knocking with "integrated solutions."
Linux is hitting the big time, as two supercomputer projects based on the freely available OS have been launched. The U.S. Department of Energy's Argonne National Laboratory recently joined forces with IBM and VA Linux Systems to build a 512-CPU Linux cluster. Called the "Chiba City Project" (after the futuristic Chiba City in William Gibson's scifi novel, Neuromancer), the system will be Argonne's most powerful supercomputer. The cluster is composed of 256 dual processor servers from VA Linux and IBM. A prepared statement from Argonne said that the "cluster installation was accomplished in a two-day 'barn raising' event, complete with banjo player. Over fifty Argonne scientists pitched in to help build the cluster, which links high-performance servers from VA Linux with advanced hardware from IBM and the latest in network interconnect hardware."
Meanwhile, Silicon Graphics (SGI) demonstrated a Linux-based supercomputer at the Supercomputing Conference '99 in Portland, Oregon. SGI is trying to sell off its Cray supercomputer division, opting instead to offer new supercomputers based on clustered 64-bit Intel IA-64 Itanium processors running Linux.
The wild speculation on Internet -- or dot-com -- company stock shares has apparently subsided, but it seems that an advertising spending spree is just getting into gear. Internet companies spent close to $2 billion in advertising in 1999. Dot-com advertisers have been a boon for radio and TV stations, not to mention billboard owners. Amazon.com alone spent $100 million in advertising for the 1999 holiday season.
Amazingly, 20 percent of the available commercials for the January 2000 Superbowl were purchased by dot-com companies. A 30-second commercial costs about $2 million. Most of these dot.com advertisers have yet to turn a profit. One Superbowl advertiser, Computer.com, spent $3 million on Super Bowl advertising and has generated less than $500,000 in revenue, according to an article in The Wall Street Journal. Anthony Perkins, editor of Red Herring and coauthor of The Internet Bubble, advises investors to sell their Internet stocks and invest in broadcast companies.
Ironically, some advertising agencies are demanding dot-com companies fork over 50 percent of the fee up front.
Engineers at the University of California, Berkeley, claim to have created the world's tiniest semiconductor transistor. Called the FenFET, the transistor uses a new gate design with improved current control and reduced current leakage, which allows the transistor to be much smaller. Unlike the conventional flat conductor gate that controls only one side of the passage through which the current flows in a transistor, the UC Berkeley design uses a fork-shaped prong that straddles both sides of the current channel.
UC Berkeley Professor Chenming Hu, the leading faculty member on the project, told DDJ that the major obstacle to shrinking the size of transistors is the gate length. If the gate length is too small, current leaks across the gate even when the transistor is in the off state. Conventional transistors have a gate length of about 180 nanometers. Professor Hu's team has successfully reduced the length to 18 nanometers using the prong design, which has solved the current leakage problem. With a factor of 10 gate-length reduction, silicon chips using these new transistors would have 100 times the capacity of chips using conventional transistors. Hu says that computer simulation tests indicate that the gate length can eventually be reduced to 9 nanometers, which would result in a 400-fold increase in chip capacity.
Professor Hu said that there are no plans to patent the device and hopes that the research will provide a "stimulant to industry for trying new approaches to semiconductor design." Professor Hu and graduate student Xuejue Huang presented their research in detail at the International Electron Devices Meeting (http:// www.ieee.org/conference/iedm).
Yale and Rice University researchers are turning to molecular computing to come up with an even smaller logic gate. The researchers have created an electronic switch the size of a single molecule. Electrical engineers from Yale and chemists from Rice joined forces to develop the molecular switch, which is described in the November 19, 1999, issue of Science magazine. Professor James Tour of Rice designed the synthetic molecule used in the switch, while Professor Mark Reed of Yale led the effort involving the electronics.
According to a prepared statement from both universities, the switch exhibits on/off ratios of close to 1000:1, compared to 50:1 for typical silicon devices. The designers expect fabrication of molecular switches to be much cheaper than semiconductors once the technology evolves into practical applications, which is estimated to be 5 to 10 years away. The research was also presented at the International Electron Devices Meeting (http://www.ieee.org/ conference/iedm).