The controversial tale of two KPU's tale
Vishnu K. Mahmud, Contributor, Jakarta, vmahmud@yahoo.com
There was a little controversy a while back, when the Indonesian General Election Committee or KPU (www.kpu.or.id) held an open tender for bids to establish the Information Technology infrastructure for the 2004 General Elections.
It was hoped that the election results could be sped up electronically by "wiring" the provinces with computers for efficient vote tallying.
However, according to ICT Watch (www.ictwatch.com), an information and communications technology non-governmental organization, the tender documents provoked many questions as they seemed to favor a particular hardware and software brand.
The documents called for hardware solutions armed with 4 x Itanium2 1.5 GHz processors, Smart Array 5i Plus controllers and/or embedded Integrated Lights-Out (iLO) that support multiple operating systems.
According to ICT Watch, only one hardware vendor in Indonesia possesses these explicit specifications.
Of course, there are many ways to draft system requirements for tenders -- although some organizations would simply draft the business processes to be implemented and call for solutions from professional IT consultants.
But they usually don't specify a particular hardware or software solution since it is the bidder's job to provide the client with an adequate solution, at the right price.
Take, for example, the system requirements of Virginia Polytechnic Institute and State University (www.vt.edu) in the U.S.
Virginia Tech (VT) wanted to create a supercomputer cluster out of a large number of off-the-shelf computer components to help attract lucrative research projects in areas such as quantum chemistry, computational biochemistry and cell cycle modeling.
Yet, they lacked the budget of government or corporate institutions, who spend tens of millions of U.S. dollars on computer hardware.
By "clustering" personal computers together, Virginia Tech hoped to have the processing power of a supercomputer at a fraction of the cost.
VT had to consider the basic needs of the system, its goals and effectiveness as well as the total cost of ownership -- including the potential location of the computers, cooling systems, electricity costs, maintenance personnel and other considerations for hosting a stable of computers to work together as one.
They thought about what the supercomputer should be able to do, instead of how. Speed and cost were the primary concerns for the university and they did not choose a hardware or software vender right off the bat as their choice had the potential to skew the system requirements of the cluster, forcing the university down costly proprietary paths without alternative solutions.
After going through various hardware vendors, Virginia Tech eventually found their solution: the Apple G5 (www.apple.com), launched in September as the world's most powerful personal computer.
They considered that the 64-bit G5 desktop computer had the potential to form one of the world's most powerful supercomputer clusters. They ordered 1,100 Dual processor G5 computers and have since been installing and connecting the PCs into one giant network.
VT's supercomputer cluster, unofficially called "Big Mac", is expected to break into the top 10 of the top 500 supercomputers in the world (www.top500.org). There are even reports that the Big Mac is ranked third at 9.55 teraflops (trillion calculations per second).
The faster supercomputer in the planet is Japan's Earth Simulator, a monster of a machine consisting of 5,000 processors, operating at 35.6 teraflops. Compared to Earth Simulator's hefty US$350 million price tag, the 2,200 processor Big Mac is a bargain at US$5.2 million.
The latest list of the top 500 is due to be released on Nov. 17 at the International Supercomputer Conference.
VT's open-ended requirements allowed the school to build a powerful supercomputer using a computer system that others have never even considered. Yet all their hard work may soon pay off with an established respectable global ranking -- and profitable research grants.
Looking back at the KPU tender process, which was recently repeated, the hardware specifications may have locked out potential vendors from offering quality solutions at a more agreeable price. Questions also remain concerning the software side.
In an open letter, I Made Wiryana (http://wiryana.pandu.org/), a Linux advocate, lamented the KPU's decision to use proprietary office software for their computers. Open Office (www.openoffice.org) is widely known to be mostly compatible with the world's leading (yet costly) productivity applications, with an agreeable price (free).
Yet Wiryana notes that the software specifications called for expensive office applications for the thousands of computers to be set up around the nation.
It's a pity that other alternatives were not considered as they may have saved the taxpayers a lot of money.