Demand-based coscheduling of parallel jobs on multiprogrammed multiprocessors

Abstract

This thesis describes demand-based c oscheduling, a new approach t o s c heduling parallel computations on multiprogrammed multiprocessors. In demand-based coscheduling, rather than making the pessimistic assumption that all the processes constituting a parallel job must be simultaneously scheduled in order to achieve good performance, information about which processes are communicating is used in order to coschedule only these; the resulting scheme is well-suited to implementation on a w orkstation cluster because it is naturally decentralized. I present an analytical model and simulations of demand-based coscheduling, an implementation on a cluster of workstations connected by a high-speed network, and a set of experimental results. An analysis of the results shows that demand-based coscheduling successfully coschedules parallel processes in a timeshared workstation cluster, signi cantly reducing the response times of parallel computations. AbstractThis thesis describes demand-based c oscheduling, a new approach t o s c heduling parallel computations on multiprogrammed multiprocessors. In demand-based coscheduling, rather than making the pessimistic assumption that all the processes constituting a parallel job must be simultaneously scheduled in order to achieve good performance, information about which processes are communicating is used in order to coschedule only these; the resulting scheme is well-suited to implementation on a w orkstation cluster because it is naturally decentralized. I present an analytical model and simulations of demand-based coscheduling, an implementation on a cluster of workstations connected by a high-speed network, and a set of experimental results. An analysis of the results shows that demand-based coscheduling successfully coschedules parallel processes in a timeshared workstation cluster, signi cantly reducing the response times of parallel computations. Acknowledgments 1Sometimes when looking through a thesis I nd that it contains a very long acknowledgments section, which more often than not proves upon examination to be a very personal document. When the writer is a friend, I enjoy reading the acknowledgments, especially if my name is mentioned. When the writer is someone I know only professionally, m y reaction is typically embarassment tinged with distaste. I am reminded of a Peanuts cartoon I read when I wa s a c hild, in which Snoopy s a w a dog riding in a car with its head out the window and its tongue apping in the breeze. Snoopy thought that if he were in a car he would never behave in so undigni ed a manner; rather, he would sit up straight and wear a seat belt. It will be clear, then, that in order to maintain a digni ed posture and spare the reader embarassment, I should like to write a brief and formal acknowledgments section, but when I look back on the work that has occupied me for most of three years, I realize that that I am so indebted to colleagues, family, and friends, that to do so would be out of the question. Thus to the reader who does not kno...

ClusterSchedSim: A Unifying Simulation Framework for Cluster Scheduling Strategies

“…• Dynamic Co-Scheduling Schemes Dynamic co-scheduling schemes are proposed to further boost the system utilization [5,1,21,15]. These schemes allocate multiple tasks (from different jobs) to a node, and leave the temporal scheduling of that node to its local CPU scheduler.…”

Coscheduling under Memory Constraints in a NOW Environment

“…In terms of locality, the local scheduler will settle for the resident set of such distributed tasks as can fit its locality. Ñ ×µ is based on the predictive technique, [14,23]. In predictive coscheduling, in contrast to the dynamic coscheduling, both send and receiving frequencies are taken into account.…”

“…Thus, coscheduling may be applied to reduce messages waiting time and make good use of the idle CPU cycles when distributed applications are executed in a cluster or NOW system. Coscheduling decisions are made taking implicit runtime information of the jobs into account, basically execution CPU cycles and communication events [12,13,14,15,16,22]. Our framework will be focused on an implicit coscheduling environment, such as scheduling the correspondents -the most recent communicated processes-in the overall system at the same time, taking into account both high message communication frequency and low penalty introduction into the delayed processes.…”

Locality-information-based scheduling in shared-memory multiprocessors

“…An additional optimization is possible, if we look at the interaction of threads on the same memory regions, by comparing their working set. In [16] TLB information is used to find cooperating processes on kernel level. For fast interaction a user-level readable TLB would be necessary, accessible as fast as the processor cache and with a fine resolution in the range of 1 kByte.…”