sai_j的博客

This J›roblem oJ “rloing things right” on a large scale is not essentially new; in a telephone central office, for exa niJ ie, a v ery large numher of operations are performccl iv lii Ie the errors lea‹1iiig to vv roii g numbers are kept well under control, though t I ey have net 1›eeu cont pletely elimiiia te‹I . This has been achieved, in part, through the use ‹if sell-checking circuits. The occasional failure that escapes routine check rug is still ‹letected 1›y' the customer an‹J iv ill, iI it persists, result in customer conipla int, v hile if it is t ransieiit it will J›roduce only occasional iv mug nu nil›crs. .\ I the same t ime the rest of the central ofhce fu nctions sa I isla ct or i I\'. lii a r1i =ita I compu ter, on the other hand, a single failure usual I y me:iris I lie complete la i l ure, in the sense that iI it is detec te‹l no more com J›utiiig can he done until the Jai lure is located ancl correcte‹l , wit i l e if it escapes rlot e‹ ti‹m their it iiival relates all suhsequent operat ions of the machine. Put in other words, in a telephone central office there are ii numlier of J:ara11e1 paths v’1iich are more or less iiiclependent of each other ; in :i digital mach ine there is usually a sinyle ton g path w'hich passes t li ro •b³' 'lie same J, of Sql many time beforethe answer is obtaine‹l.