SI
SI
discoversearch

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 : Intel Corporation (INTC)
INTC 50.59+4.9%Feb 6 9:30 AM EST

 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: Tenchusatsu who wrote (134447)5/8/2001 7:44:32 PM
From: Saturn V  Read Replies (1) of 186894
 
Ref <Without error detection and correction, how did SUN's software figure out good data from bad?>

I got this info second hand from one customer who was involved .

To figure out the problem on large servers with multiple processors, SUN implemented redundancy, ie one processor module compares its results with another module which is supposed to have the identical data and instruction set. When the results did not match up, the system would halt and try to dump the status info. However this dump was not implemented cleanly,and the system would crash.

To me it sounded that SUN ended up debugging this problem in the field with the cooperation and assistance of several customers, and it took several months of painful data gathering and crashes before the light dawned at SUN. Now the reasons for the NDA become more reasonable.

Once understood the fix was to replace the processor board with new ones, which were less prone to soft errors.[Watsonyouth reported that SUN got the better L2 memory modules from IBM]. Since this was expensive and slow, SUN dragged its feet, and demanded a large amount of history of crashes before SUN would agree to replace the processor boards.
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext