Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Silicon Graphics, Inc. (SGI) -- Ignore unavailable to you. Want to Upgrade?

To: Jeff-P who wrote (3932)	1/3/1998 12:19:00 AM
From: Patrick Gainer	Read Replies (1) \| Respond to of 14451

"SGI itself learned a hard lesson with its Power Challenge system, and has since turned to a directory-based shared memory system in the Origin 2000 series. Proposing that the solution to this problem is straightforward, or only a matter of implementing semaphores, is naive." I think you've muddled a bunch of ideas together in your response. 1, The scale-ability discussion so far has not involved hardware issues, only software issues. This point is relevant because we are talking about OSes. Hardware and software scale-ability are related but different issues. 2, It wasn't a "hard" lesson. They were different machines with different design points, designed 5 years apart, and designed with different scale-ability goals. And anyway, it wasn't the cache coherence algorithms which limit the Challenge's scale-ability, it was ultimately the bus bandwidth available. 3, Nobody proposed implementing semaphores as a solution to anything. You weren't paying attention and I wasn't being naive. 4, I said that going on and on about OS scaling is dangerous because things can change very rapidly. Scale-ability can easily change drastically from one release to another. 5, I went further and said making software scale is a very well understood problem. Mr Cousins agrees. All that is required is to focus the engineers on the problem. What the SGI folks perhaps fail to appreciate is that MS has not yet really cared too much about scaling beyond 4 cpus because up until recently, NT machines with more than 4 cpus were rare. When they care, I would be willing to bet it starts to scale. 6, I mentioned that TPC-C benchmarks using NT have recently been released using 8 cpu machines. I only make this point to illustrate there is progress on the NT front. You might argue TPC-C isn't a useful comparison but it is an audited benchmark and the only other things I've seen mentioned in this forum are personal anecdotes by people who work at SGI. No conflict of interest there, is there? Pat