Original URL: https://www.theregister.com/1999/06/10/linux_disties_clash_over_powerpc/

Linux disties clash over PowerPC clustering quality

Is PowerPC better for parallel computing than Intel, Alpha or not?

By Tony Smith

Posted in On-Prem, 10th June 1999 15:32 GMT

The Mac Linux community was in a some confusion today when it emerged that attempts to show the Linux on the PowerPC was an ideal platform for parallel computing systems might not be all they were cracked up to be. TerraSoft Solutions (TSS) a couple of days ago announced Black Lab Linux, a version of the company's Champion Server Linux distribution for cluster computing. TSS claims Black Lab Linux offers a "robust, low cost and low power consumption, high performance advanced computing solution". Essentially, the Black Lab business designs and specifies clusters of PowerPC-based machines running Champion Server. Linux on PowerPC, a company consultant claimed, was "a viable and competitive solution" for parallel computing. But today, a spokesman for rival Linux distributor LinuxPPC, claimed TerraSoft had got it wrong. Cited by leading Mac information service, MacInTouch, he said a "group of researchers at the Distributed and High-Performance Computing (DHPC) Group at University of Adelaide in Australia published a paper finding that Linux on PPC is a poor choice for clustering". The DHPC group's findings, published in March, compared a cluster of iMacs, a Beowulf-style Pentium-based PC cluster and a network of DEC Alpha workstations. In benchmarks, the iMac system didn't come out too well compared with the others. "The group used iMacs and an older version of LinuxPPC, but the results are clear: Intel and Alpha clusters massively outperformed the iMac cluster," said the LinuxPPC spokesman. So, who is correct? Well, the DHPC group did use an old version of LinuxPPC, whereas Black Lab Linux is based on a far more recent release of the Linux kernel. And, Black Lab doesn't claim to be better than Alpha or Pentium solutions, merely that it consumes less power. Still, that doesn't necessarily make it a "viable" or "competitive" solution, either. We were unable to download the University of Adelaide report at the time of posting to confirm their findings. However, a follow-up survey, using more recent versions of both LinuxPPC and Champion Server, would certainly go some way to clear up the issue. ®