# Abstract

$B9-0h%M%C%H%o!<%/$K@\B3$5$l$?7W;;;q8;$r(B1$B$D$N7W;;4pHW$H$7$FDs6!$9$k(B Grid$B%7%9%F%$,>-Mh$N9b@-G=7W;;%"%W%j%1!<%7%g%s$N7W;;4pHW$H$7$F(B $BCmL\$5$l$F$$k!%(B GridB%7%9%F%H7F!%M%C%H%o!<%/>eG7W;;;q8;HHbK%5!<%S%9NDs6!r(B B2DG=K9k(BNetwork-enabled ServerB!J(BNESB!K,J#?tDs0F5lF$$$k!%(B NES$B$O0lHLE*$K%/%i%$%"%s%H!&%5!<%P7?%"!<%-%F%/%A%c$H$J$C$F$*$j!$(B $BJ,;6$7$?(BGrid$B7W;;;q8;>e$K%5!<%P$rMQ0U$9$k!%(B $B$7$+$7$J$,$i!$J#?t%5!<%P!$J#?t$N%/%i%$%"%s%H$rA[Dj$7$?(BGrid$B$N(B $B%9%1%8%e!<%j%s%0$K4X$9$k5DO@$,==J,$K9T$o$l$F$$J$$!%(B $B$^$?!$>-Mh$N(BNES$B%7%9%F%1?MQ$G$N2]6b$KH<$$!(BGridB%f!<%6O%8%g%V(B B.%3%9%HN;q8;72rMxMQ7F5,Dj;~4VFbK(B B=hM}r=*N;5;k3HrMW5a9kh&KJk!%(B BK\9FGO%G%C%I%i%%s%9%1%8%e!<%j%s%0KCeL\7!$$=$N@-G=FC@-$r(BGrid $B$N%9(B $B%1%8%e!<%j%s%02A%7%9%F%(BBricks$B$r3HD%$7$FD4::$7$?!%(B $B$^$:!$J#?t%5!<%P!$J#?t%/%i%$%"%s%H$rA[Dj$7$?!$%G%C%I%i%$%s%9%1%8%e!<%j%s%0(B $B%"%k%4%j%:%$r>R2p$9$k$H$H$b$K!$$=$N%"%k%4%j%:%$N@-G=$r9b$a$k%a(B $B%+%K%:%!$(BLoad Correction$B$H(BFallback$B$rDs0F$9$k!%(B $B2A$h$j!$(BGrid$B>e$G$N(B NES$B%7%9%F%$N%G%C%I%i%$%s%9%1%8%e!<%j%s%0$NM-8z@-$r<(\$9!%(B
The Computational Grid is a promising platform for the deployment of high-performance computing applications. A number of projects have addressed the idea of software as a service on the network. These systems usually implement client-server architectures with many servers running on distributed Grid resources and have commonly been referred to as Network-enabled servers (NES). An important question is that of {\em scheduling} in this multi-client multi-server scenario. Note that in this context most requests are computationally intensive as they are generated by high-performance computing applications. The Bricks simulation framework has been developed and extensively used to evaluate scheduling strategies for NES systems. In this paper we discuss a deadline scheduling strategy that is appropriate for the multi-client multi-server case. We propose a simple deadline scheduling algorithm and then augment it with Load Correction and Fallback mechanisms which could improve the performance of the algorithm in our context. We then give simulation results and draw conclusion on the value and feasibility of deadline scheduling for NES systems on the Grid.