There is strength in numbers

Geniuses and mortals alike have suggested that humans use only a small fraction of their brainpower. Computers also seem to suffer from this same affliction, with studies showing them to be idle 90 per cent of the time. Though there is little hope for humans, computers could turn to LSF to help overcome their slacker ways.

Platform Computing Corp., a Markham Ont.-based application resource management software company, recently released Version 4.0 of their flagship product, Load Sharing Facility. LSF allows users to harness all available computing resources for jobs that require maximum computing power, from CAD renderings to mapping the human genome.

Today’s projects are dramatically increasing in their complexity, according to David Wilmering, Platform’s senior group product manager. One way to increase corporate efficiency is to make more productive use of a company’s computing resources, he added.

“The more analysis I can do in a finite time, the better my market position, the better my market position [and] I win,” Wilmering said.

Platform helps companies maximize the efficient use of their computing power by creating a network where workers can access the power of workstations and servers to supplement the limited power they have on their desktops.

By distributing the workload through load sharing on mixed platform computing environments, LSF 4.0 can harness all available computing power, according to the company.

“We can get the best out of your infrastructure…we can take that unwashed mass of computing capacity that is not used today, we can align it with your business policies,” Wilmering said.

For Rick Lantaigne, Unix project manager at Nortel Networks Corp. in Nepean, Ont., LSF has been a real time saver. It has also cut down on the need to baby-sit jobs.

“This is what the guys like. They submit it (any job) to the cue and they can just check on the status of it through their e-mail, or they can log in and check on the status as opposed to having to watch it,” he said. “It helps make better use of the high-end compute machines that we have.”

He said before using LSF, the engineers had to continuously monitor job progress. Now they can submit four or five variations of the same work at one time and LSF will farm it out to the computers, Lantainge said.

Lantainge’s group uses LSF for Application Specific Integrated Circuit design and has a pool of machines that the engineers can use to submit work. He said the engineers send hundreds of jobs to the computers on Friday afternoon and when they return on Monday the jobs are done.

As added security, LSF will move a job to another machine if a crash occurs, according to Wilmering. Even if the master computer crashes it is not a problem since LSF will automatically and seamlessly choose another to be the master, he added. A master computer can cluster around 3,000 servers and clusters themselves can be joined.

Since LSF allows for almost limitless network size, it can help address the issue of scalability that has plagued Windows NT. The National Center for Super Computing at the University of Illinois at Urbana-Champaign has the world’s largest NT cluster and it runs on LSF, Wilmering said.

Analyst Dan Kusnetsky, vice-president of systems software for Framingham, Mass.-based IDC, senses that Platform’s work is cut out for them as they shift from the high performance and technical computing community, where they are well established, to the Internet.

“Platform has a very good product but now they are facing some new marketing challenges from a number of different places…as they move into other markets they, in essence, have to start over again,” he said.

“I have a sense if they take the strength of this product…they will build a strong reputation because the quality of the product is quite good,” he concluded.

Version 4.0 has upgraded several portions of the software including easier installation, so different host types and architectures can be done simultaneously, and increased customization for Analyser, which assesses cluster performance.

Platform LSF 4.0 ( is priced starting around US$400 per CPU and is available for Unix, Linux and NT platforms.

Platform in Markham, Ont., can be reached at 1-877-528-3676.