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.

Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?

To: Mani1 who wrote (82548)	12/9/1999 9:10:00 PM
From: Mani1	Read Replies (2) \| Respond to of 1573050

This, link from JC, is cool! zdnet.com New Perpetua server built to withstand attacks By Sonia R. Lelii, PC Week Online December 3, 1999 1:54 PM ET Call it the B-17 Bomber of servers because it never crashes. Startup Patmos International Corp. this week introduced an eight-node server cluster, compacted into an 84-inch-high machine, that uses artificial intelligence to rebalance the data load among the servers when one or more go down. The device, called Perpetua, starts with eight independent servers networked with switching and routing capabilities inside a single box. Each server, equipped with a 450MHz Advanced Micro Devices Inc. K6 processor and one hard drive, can run its own Unix, Linux, Windows or NetWare operating system. How it stacks up The servers are connected to two special-purpose networking compartments, called Limbix, that run Patmos' proprietary Parpois network management software. Parpois automatically manages the workload between the various nodes as well as fails over processing duties between servers when one fails. Each Limbix is equipped with a 750MHz AMD Athlon processor. A third special-purpose backup compartment, called the nMime, contains eight hard drives that mirror the drives on the eight server nodes. The Limbix and nMime work in tandem, when a server node fails, to make sure that applications continue running. When the defective node restarts, the Limbix continues to do the processing while the nMime fills in the data it missed while it was out of commission. The eight servers, two Limbix and nMime are interconnected via an internal, full-duplex Fibre Channel network that uses a 200MHz bus and boasts a 1.065G-bps transfer rate. A 'virtual vault' The redundancy creates a virtual vault that keeps IT operations up and running, said James Gatzka, co-founder and CEO of Patmos in Ocean City, Md. "This machine never, ever goes down," Gatzka said. "You have to go through 10 processes before you can take this thing down." Perpetua also has artificial intelligence that can recognize patterns and learn from mistakes, Gatzka said. "If it sees that something has happened three or four times, it speculates on what is causing it," Gatzka said. Perpetua was also engineered to withstand attacks from outside forces. It has two battery-powered power supplies, a backup generator and a cabinet can withstand the force of an earthquake measuring 7.5 on the Richter scale, officials said. Pricing for Perpetua, which is available now, begins at around $150,000 to $200,000. One of the first customers to sample Perpetua is Honeywell Security Products (formerly Westinghouse Security Electronics). The Fremont, Calif., company purchased a four-node box with one Limbix system and is running Red Hat Linux 6.0. Jeri Donn, director of software development at Honeywell, said she had just installed the system. "On paper, at least, we are believers. Now we have to see how it works," Donn said. "The way it's designed, if we pull one component out, it will continue to work. In terms of specifications, I haven't seen anything that comes close to this. And at this kind of price, people will go for it."

To: Mani1 who wrote (82548)	12/9/1999 9:12:00 PM
From: Elmer	Respond to of 1573050

Re: "Why don't you just admit that you were wrong in bad mouthing AMD because of that P=VI? I agree with Kash that, after Intel's numbers was shown you ignored that topic which brought up this whole argument / discussion. Everyone on the thread knows that you made a mistake, bad mouthed AMD, and not once admitted you were wrong when it has clearly been shown otherwise here." Mani, I don't feel I intentionally avoided the topic. Please understand my position. I was carrying on 4 separate discussions with 4 separate people. You, Bill, Petz, and Kash. 3 of you were interested in following a different line by a sub-thread which kept forking again and the 4th was only interested in trashing me. Could you keep perfect continuity under the circumstances? I don't feel the need to waste a lot of energy on someone who has repeatedly called me a liar and has real trouble following technical discussions. So it wasn't the question I was avoiding, but wasting time explaining everything to Kash. I just didn't want to get lost down a rat-hole correcting his continuous misunderstandings and imagination. So back to your question. Does Intel use the same method for deriving their power numbers as AMD and if so why didn't I criticize Intel as well? Apparently Intel does use the same method. The one difference I see is that Intel tells you they are using nominal Vcc to calculate power while AMD tells you that you must use MAX_VCC yet publishes a number based on nom Vcc. EP