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 34.50+2.6%Nov 21 9:30 AM EST
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: Tony Viola who wrote (108034)8/22/2000 8:24:25 PM
From: Tenchusatsu  Read Replies (1) of 186894
 
Tony, <You're going to put 72 PCs together to make a server?>

To quote an article I just linked to earlier today:

zdnet.com

The results provide continuing evidence to IT that scaling up an enterprise's database back-end by using shared-nothing clustering techniques on commodity hardware may provide better performance at much lower costs than buying larger symmetric multi-processing servers. The key constraint is that an organization's data set and queries must be suited for partitioning between servers.

In other words, large numbers of cheap boxes will provide better performance and price/performance than small numbers of expensive boxes. In this case, the "cheap box" is the 4-way Xeon, and the "expensive box" is something like a Sun Enterprise 6500 server.

I think Pete is just extending that argument. If you can get better performance out of large numbers of SMP Xeons, why not go even further and use even larger numbers of uniprocessor systems like Athlon? Hey, if someone can actually pull off something like Pete is suggesting, then more power to them.

I believe, however, that 4-way SMP boxes currently represent the sweet spot in server clusters. Going further would only increase clustering overhead and will probably create a big drop-off in performance and scalability.

By the way, IBM actually used 8-way Xeon servers (i.e. Profusion) in their record-setting TPC-C benchmark. For some reason (probably lack of scalability), they chose to populate only half of the processor slots in each server.

Tenchusatsu
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  

HomeMailHotSubjectMarksPeopleMarksKeepersSettings
Terms Of UseContact UsCopyright/IP PolicyPrivacy PolicyAbout UsFAQAdvertise on SI
© 2025 Knight Sac Media.  Data provided by Twelve Data, Alpha Vantage, and CityFALCON News