Improvements in multiprocessor system design

Three Factors have l i m i t e d the commerc i a l success of m u l t i p r o c e s s o r systems: en t rg l e v e l cost , range of performance, and ease of a p p l i c a t i o n . Recent lg , the con f luence o f s e v e r a l streams o f development in computer system design has removed these l i m i t i n g Factors making p o s s i b l e a new c lass of m u l t i p r o c e s sot systems based on VLSI components o f f e r i n g mainframe performance a t m i c ro computer p r i ces in t r a d i t i o n a l t i m e s h a r ing env i ronments and ho ld ing the promise of supercomputer performance a t onehundredth the cost in s p e c i a l i z e d a p p l i ca t i ons . M u l t i p r o c e s s o r s have been of i n t e r e s t to computer s c i e n t i s t s and des igners from the t ime of the very F i r s t e l e c t r o n i c computers to the present . A number of commerc ia l lg a v a i l a b l e m u l t i p r o c e s s o r s have appeared over t h a t span and some have enjoyed modest success. Three Fact o r s have l i m i t e d the commercial success of m u l t i p r o c e s s o r sgstems: en t r y l e v e l cost , range of performance, and ease of a p p l i c a t i o n . In f i r s t , second and t h i r d g e n e r a t i o n t e c h n o l o g i e s , each processor in a dual processor sgstem r e q u i r e d as much ~uppor t ing power, packaging, c o o l ing and s e r v i c e as two independent un ip rocessor systems making the cost o f a minimum m u l t i p r o c e s s o r sgstem high and s u s c e p t i b l e to p r i c e p e r f o r m a n c e e c l i p s e bg r e l a t i v e l y smal l advances in t e c h n o l ogy f o r un iprocessors . When a p p l i e d to m u l t i p r o c e s s o r s , the hardware and so f tware system design p r a c t i c e s which y i e l d good un ip rocessor performance lead to performance b o t t l e n e c k s due to memory bandwidth and system s e r v i c e resource s t a r v a t i o n . These b o t t l e n e c k s l i m i t the range of c o s t e f f e c t i v e m u l t i p r o c e s s i n g to a t most f ou r processors. E x p l o i t a t i o n oF concu r ren t program execu t i on by a p p l i c a t i o n s so f tware r e q u i r e s under s tand ing of and techno logy For problem decompos i t ion , load d i s t r i b u t i o n , sgnc h r o n i z a t i o n and hazard avoidance. These ave d i f f i c u l t areas, t r a d i t i o n a l l y l e f t to p r a c t i t i o n e r s of r e a l t i m e mult i t a s k i n g , f o r which the re have been few t o o l s and l i t t l e o p e r a t i n g sgstem and language support . Recent lg , the con f luence of seve ra l streams of development in computer ~gstem design has removed these l i m i t i n g Factors making p o s s i b l e a new c lass of m u l t i p r o c e s s o r systems based on VLSI components o f f e r i n g mainframe p e r f o r mance a t microcomputer p r i c e s in t r a d i t i o n a l t imesha r i ng env i ronments and ho ld ing the promise o f supercomputer performance a t one-hundredth the cost in s p e c i a l i z e d a p p l i c a t i o n s . Recogniz ing t h i s p o t e n t i a l , a number of computer manufac tu rers are reduc ing these developments to p r a c t i c e in the f o r o£ m u l t i p r o c e s s o r computer sgstems.