DOCSLIB.ORG

Job Scheduling Strategies for Networks of Workstations

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Job Scheduling Strategies for Networks of Workstations Job Scheduling Strategies for Networks of Workstations 1 1 2 3 B B Zhou R P Brent D Walsh and K Suzaki 1 Computer Sciences Lab oratory Australian National University Canb erra ACT Australia 2 CAP Research Program Australian National University Canb erra ACT Australia 3 Electrotechnical Lab oratory Umezono sukuba Ibaraki Japan Abstract In this pap er we rst intro duce the concepts of utilisation ra tio and eective sp eedup and their relations to the system p erformance We then describ e a twolevel scheduling scheme which can b e used to achieve go o d p erformance for parallel jobs and go o d resp onse for inter active sequential jobs and also to balance b oth parallel and sequential workloads The twolevel scheduling can b e implemented by intro ducing on each pro cessor a registration oce We also intro duce a lo ose gang scheduling scheme This scheme is scalable and has many advantages over existing explicit and implicit coscheduling schemes for scheduling parallel jobs under a time sharing environment Intro duction The trend of parallel computer developments is toward networks of worksta tions or scalable paral lel systems In this typ e of system each pro cessor having a highsp eed pro cessing element a large memory space and full function ality of a standard op erating system can op erate as a standalone workstation for sequential computing Interconnected by highbandwidth and lowlatency networks the pro cessors can also b e used for parallel computing To establish a truly generalpurp ose and userfriendly system one of the main problems is to provide users with a single system image By adopting the technique of dis tributed shared memory for example we can provide a single addressing space for the whole system so that communication for transferring data b etween pro cessors is completely transparent to the client programs In this pap er we discuss another very imp ortant issue relating to the provision of single system image that is eective job scheduling strategies for b oth sequential and paral lel processing on networks of workstations Many job scheduling schemes have b een intro duced in the literature and some of them implemented on commercial parallel systems These scheduling schemes for parallel systems can b e classied into either space sharing or time sharing or a combination of b oth With space sharing a system is partitioned into subsystems each containing a subset of pro cessors There are b oundary lines laid b etween subsystems and so only pro cessors of the same subsystem can b e co ordinated to solve problems assigned to that subsystem During the computation each subsystem is allo cated only for a single job at a time The space partition can b e either static or adaptive With static partitioning the system conguration is determined b efore the system starts op erating The whole system has to b e stopp ed when the system needs to b e recongured With adaptive partitioning pro cessors in the system are not divided b efore the computation When a new job arrives a job manager in the system rst lo cates idle pro cessors and then allo cates certain numb er of those idle pro cessors to that job according to some pro cessor allo cation p olicies eg those describ ed in Therefore the b oundary lines are drawn during the computation and will disapp ear after the job is terminated Normally the static partitioning is used for very large systems while the adaptive partitioning is adopted in systems or subsystems of small to medium size One disadvantage of space partitioning is that short jobs can easily b e blo cked by long ones for a long time b efore b eing executed However in practice short jobs usually demand a short turnaround time To alleviate this problem jobs can b e group ed into classes and a sp ecial treatment will b e given to the class of short jobs However it can only partially solve the problem Thus time sharing needs to b e considered Many scheduling schemes for timesharing of a parallel system have b een prop osed in the literature They may b e classied into two basic typ es The rst one is local scheduling With lo cal scheduling there is only a single queue on each pro cessor Except for higher or lower priorities b eing given pro cesses asso ciated with parallel jobs are not distinguished from those asso ciated with sequential jobs The metho d simply relies on existing lo cal schedulers on each pro cessor to schedule parallel jobs Thus there is no guarantee that the pro cesses b elonging to the same parallel job can b e executed at the same time across the pro cessors When many parallel programs are simultaneously running on a sys tem pro cesses b elonging to dierent jobs will comp ete for resources with each other and then some pro cesses have to b e blo cked when communicating or syn chronising with nonscheduled pro cesses on other pro cessors This eect can lead to a great degradation in overall system p erformance One metho d to alleviate this problem is to use twophase blo cking which is also called implicit coscheduling in In this metho d a pro cess waiting for communication spins for some time in the hop e that the pro cess to b e communicated with on the other pro cessor is also scheduled and then blo cks if a resp onse has not b een received The rep orted exp erimental results show that for parallel workloads this scheduling scheme p erforms b etter than the simple lo cal scheduling However the problem is that the scheduling p olicy is based on communication require ments Then it tends to give sp ecial treatment to jobs with a high frequency of communication demands The p olicy is also indep endent of service times The p erformance of parallel computation is thus unpredictable The second typ e of scheduling schemes for time sharing is coscheduling or gang scheduling which may b e a b etter scheme in adopting shortjobrst p olicy Using this metho d a numb er of parallel programs is allowed to enter a service queue as long as the system has enough memory space The pro cesses of the same job will run simultaneously across the pro cessors for only certain amount of time which is called scheduling slot When a scheduling slot is ended the pro cessors will contextswitch at the same time to give the service to pro cesses of another job All programs in the service queue take turns to receive the service in a co ordinated manner across the pro cessors Thus programs never interfere with each other and short jobs are likely to b e completed more quickly There are also certain drawbacks asso ciated with coscheduling A signicant one is that it is designed only for parallel workloads For networks of workstations we need an eective scheduling strategy for b oth sequential and parallel pro cessing The simple coscheduling technique is not a suitable solution The future networks of workstations should provide a programmingfree en vironment to general users By providing a variety of highp erformance com puting libraries for a wide range of applications plus userfriendly interfaces for the access to those libraries parallel computing will no longer b e considered just as clients sp ecial requests but b ecome a natural and common phenomenon in the system Along with many other critical issues therefore highly eective job management strategies are required for the system to meet various clients requirements and to achieve high eciency of resource utilisation Because of the lack of ecient job scheduling strategies most networks of workstations are currently used exclusively either as an MPP for pro cessing parallel batch jobs or as a group of separate pro cessors for interactive sequential jobs The p oten tial p ower of this typ e of system are not exploited eectively and the system resources are not utilised eciently under these circumstances In this pap er we discuss some new ideas for eectively scheduling b oth se quential and parallel workloads on networks of workstations To achieve a de sired p erformance for a parallel job on a network of workstations with a variety of comp etitive background workloads it is essential to provide a sustained ra tio of CPU utilisation to the asso ciated pro cesses on each pro cessor to allo cate more pro cessors to the job if the assigned utilisation ratio is small and then to co ordinate the execution across the pro cessors We rst intro duce the concepts of utilisation ratio and eective speedup and their relations to the system p er formance in Section In this section we also argue that b ecause the resources in a system are limited one cannot guarantee every parallel job to have a sus tained CPU utilisation ratio in a time sharing environment One way to solve the problem is that we give short jobs sustained utilisation ratios to ensure a short turnaround time while to each large job we allo cate a large numb er of pro cessors and assign a utilisation ratio which can vary in a large range according to the current system workload so that small jobs will not b e blo cked and the resource utilisation can b e kept high we then present in Section a twolevel scheduling scheme which can b e used to achieve go o d p erformance for paral lel jobs and go o d resp onse for interactive sequential jobs and also to balance b oth parallel and sequential workloads The twolevel scheduling can b e imple mented by intro ducing on each pro cessor a registration oce which is describ ed in Section We discuss a scalable coscheduling scheme loose gang schedul ing in Section This scheme requires b oth global and lo cal job managers It is scalable b ecause the coscheduling
Load more

Recommended publications
  • Three Level Scheduling
    Uniprocessor Scheduling •Basic Concepts •Scheduling Criteria •Scheduling Algorithms Three level scheduling 2 1 Types of Scheduling 3 Long- and Medium-Term Schedulers Long-term scheduler • Determines which programs are admitted to the system (ie to become processes) • requests can be denied if e.g. thrashing or overload Medium-term scheduler • decides when/which processes to suspend/resume • Both control the degree of multiprogramming – More processes, smaller percentage of time each process is executed 4 2 Short-Term Scheduler • Decides which process will be dispatched; invoked upon – Clock interrupts – I/O interrupts – Operating system calls – Signals • Dispatch latency – time it takes for the dispatcher to stop one process and start another running; the dominating factors involve: – switching context – selecting the new process to dispatch 5 CPU–I/O Burst Cycle • Process execution consists of a cycle of – CPU execution and – I/O wait. • A process may be – CPU-bound – IO-bound 6 3 Scheduling Criteria- Optimization goals CPU utilization – keep CPU as busy as possible Throughput – # of processes that complete their execution per time unit Response time – amount of time it takes from when a request was submitted until the first response is produced (execution + waiting time in ready queue) – Turnaround time – amount of time to execute a particular process (execution + all the waiting); involves IO schedulers also Fairness - watch priorities, avoid starvation, ... Scheduler Efficiency - overhead (e.g. context switching, computing priorities,
    [Show full text]
  • Balancing Efficiency and Fairness in Heterogeneous GPU Clusters for Deep Learning
    Balancing Efficiency and Fairness in Heterogeneous GPU Clusters for Deep Learning Shubham Chaudhary Ramachandran Ramjee Muthian Sivathanu [email protected] [email protected] [email protected] Microsoft Research India Microsoft Research India Microsoft Research India Nipun Kwatra Srinidhi Viswanatha [email protected] [email protected] Microsoft Research India Microsoft Research India Abstract Learning. In Fifteenth European Conference on Computer Sys- tems (EuroSys ’20), April 27–30, 2020, Heraklion, Greece. We present Gandivafair, a distributed, fair share sched- uler that balances conflicting goals of efficiency and fair- ACM, New York, NY, USA, 16 pages. https://doi.org/10. 1145/3342195.3387555 ness in GPU clusters for deep learning training (DLT). Gandivafair provides performance isolation between users, enabling multiple users to share a single cluster, thus, 1 Introduction maximizing cluster efficiency. Gandivafair is the first Love Resource only grows by sharing. You can only have scheduler that allocates cluster-wide GPU time fairly more for yourself by giving it away to others. among active users. - Brian Tracy Gandivafair achieves efficiency and fairness despite cluster heterogeneity. Data centers host a mix of GPU Several products that are an integral part of modern generations because of the rapid pace at which newer living, such as web search and voice assistants, are pow- and faster GPUs are released. As the newer generations ered by deep learning. Companies building such products face higher demand from users, older GPU generations spend significant resources for deep learning training suffer poor utilization, thus reducing cluster efficiency. (DLT), managing large GPU clusters with tens of thou- Gandivafair profiles the variable marginal utility across sands of GPUs.
    [Show full text]
  • Rotating Between Scheduling Algorithms
    CARNEGIE MELLON UNIVERSITY - 15418: PARALLEL COMPUTER ARCHITECTURE AND PROGRAMMING 1 sdgOS: Rotating Between Scheduling Algorithms Valerie Choung and Samuel Damashek Abstract| We extended the SouperDamGoodOS (sdgOS) from 15-410 to support symmetric multiprocessing (SMP). Additionally, we created a mechanism for a process to select which scheduling mode it would prefer to run under. The two scheduling modes currently supported are a normal round-robin scheduling algorithm with thread-level granularity, and the other is an approximation of gang scheduling. In this paper, we discuss the implementation details our operating system and also analyze the performance of some sample programs under both scheduling algorithms. Keywords|Scheduling, Gang scheduling, OS, Autogroup. F 1 Introduction very similar to gang scheduling, except that if a processor cannot find more work from the cur- A very simple scheduling algorithm (which rently active process, it will steal work from an- we call the \normal scheduling mode", or NSM) other process's work queue. may treat all threads equally, and rotate through Our project builds on the SouperDamGoodOS threads in a round-robin fashion. However, this (sdgOS), which runs on an x86 machine support- introduces process/task-level fairness issues: if one ing a PS/2 keyboard. This environment can be process creates many threads, threads in that pro- simulated on Simics. Our study primarily has cess would take up more processor time slices in three parts to it: a given time frame when compared against other, less thread-intensive processes. 1) Implement symmetric multi-processing Gang scheduling is one way to ensure that a (SMP) process experiences a reasonable number of clock 2) Support scheduling algorithm rotations ticks before the process is switched away from to 3) Implement as syscall to switch a process be run at a later time.
    [Show full text]
  • CPU Scheduling
    CPU Scheduling CS 416: Operating Systems Design Department of Computer Science Rutgers University http://www.cs.rutgers.edu/~vinodg/teaching/416 What and Why? What is processor scheduling? Why? At first to share an expensive resource ± multiprogramming Now to perform concurrent tasks because processor is so powerful Future looks like past + now Computing utility – large data/processing centers use multiprogramming to maximize resource utilization Systems still powerful enough for each user to run multiple concurrent tasks Rutgers University 2 CS 416: Operating Systems Assumptions Pool of jobs contending for the CPU Jobs are independent and compete for resources (this assumption is not true for all systems/scenarios) Scheduler mediates between jobs to optimize some performance criterion In this lecture, we will talk about processes and threads interchangeably. We will assume a single-threaded CPU. Rutgers University 3 CS 416: Operating Systems Multiprogramming Example Process A 1 sec start idle; input idle; input stop Process B start idle; input idle; input stop Time = 10 seconds Rutgers University 4 CS 416: Operating Systems Multiprogramming Example (cont) Process A Process B start B start idle; input idle; input stop A idle; input idle; input stop B Total Time = 20 seconds Throughput = 2 jobs in 20 seconds = 0.1 jobs/second Avg. Waiting Time = (0+10)/2 = 5 seconds Rutgers University 5 CS 416: Operating Systems Multiprogramming Example (cont) Process A start idle; input idle; input stop A context switch context switch to B to A Process B idle; input idle; input stop B Throughput = 2 jobs in 11 seconds = 0.18 jobs/second Avg.
    [Show full text]
  • CPU Scheduling
    10/6/16 CS 422/522 Design & Implementation of Operating Systems Lecture 11: CPU Scheduling Zhong Shao Dept. of Computer Science Yale University Acknowledgement: some slides are taken from previous versions of the CS422/522 lectures taught by Prof. Bryan Ford and Dr. David Wolinsky, and also from the official set of slides accompanying the OSPP textbook by Anderson and Dahlin. CPU scheduler ◆ Selects from among the processes in memory that are ready to execute, and allocates the CPU to one of them. ◆ CPU scheduling decisions may take place when a process: 1. switches from running to waiting state. 2. switches from running to ready state. 3. switches from waiting to ready. 4. terminates. ◆ Scheduling under 1 and 4 is nonpreemptive. ◆ All other scheduling is preemptive. 1 10/6/16 Main points ◆ Scheduling policy: what to do next, when there are multiple threads ready to run – Or multiple packets to send, or web requests to serve, or … ◆ Definitions – response time, throughput, predictability ◆ Uniprocessor policies – FIFO, round robin, optimal – multilevel feedback as approximation of optimal ◆ Multiprocessor policies – Affinity scheduling, gang scheduling ◆ Queueing theory – Can you predict/improve a system’s response time? Example ◆ You manage a web site, that suddenly becomes wildly popular. Do you? – Buy more hardware? – Implement a different scheduling policy? – Turn away some users? Which ones? ◆ How much worse will performance get if the web site becomes even more popular? 2 10/6/16 Definitions ◆ Task/Job – User request: e.g., mouse
    [Show full text]
  • Flexible Coscheduling: Mitigating Load Imbalance Title: and Improving Utilization of Heterogeneous Resources
    LA-UR-02-6505 Approved for public release; distribution is unlimited. Flexible CoScheduling: Mitigating Load Imbalance Title: and Improving Utilization of Heterogeneous Resources Author(s): Eitan Frachtenberg, Dror G. Feitelson, Fabrizio Petrini and Juan Fernandez Submitted to: http://lib-www.lanl.gov/cgi-bin/getfile?00937216.pdf Los Alamos National Laboratory, an affirmative action/equal opportunity employer, is operated by the University of California for the U.S. Department of Energy under contract W-7405-ENG-36. By acceptance of this article, the publisher recognizes that the U.S. Government retains a nonexclusive, royalty- free license to publish or reproduce the published form of this contribution, or to allow others to do so, for U.S. Government purposes. Los Alamos National Laboratory requests that the publisher identify this article as work performed under the auspices of the U.S. Department of Energy. Los Alamos National Laboratory strongly supports academic freedom and a researcher's right to publish; as an institution, however, the Laboratory does not endorse the viewpoint of a publication or guarantee its technical correctness. FORM 836 (10/96) Flexible CoScheduling: Mitigating Load Imbalance and Improving Utilization of Heterogeneous Resources Eitan Frachtenberg1,2, Dror G. Feitelson2, Fabrizio Petrini1 and Juan Fernandez1 1CCS-3 Modeling, Algorithms, and Informatics Group 2School of Computer Science and Engineering Computer and Computational Sciences (CCS) Division The Hebrew University, Jerusalem, Israel Los Alamos National Laboratory (LANL) [email protected] {eitanf,fabrizio,juanf}@lanl.gov Abstract Performance Computing (HPC) software can be modeled using the bulk-synchronous parallel (BSP) model. In this Fine-grained parallel applications require all their pro- model a computation involves a number of supersteps, each cesses to run simultaneously on distinct processors to having several parallel computational threads that synchro- achieve good efficiency.
    [Show full text]
  • Processor Scheduling
    Background • The previous lecture introduced the basics of concurrency Processor Scheduling – Processes and threads – Definition, representation, management • We now understand how a programmer can spawn concurrent computations • The OS now needs to partition one of the central resources, the CPU, between these concurrent tasks 2 Scheduling Scheduling • The scheduler is the manager of the CPU • Threads alternate between performing I/O and resource performing computation • In general, the scheduler runs: • It makes allocation decisions – it chooses to run – when a process switches from running to waiting certain processes over others from the ready – when a process is created or terminated queue – when an interrupt occurs – Zero threads: just loop in the idle loop • In a non-preemptive system, the scheduler waits for a running process to explicitly block, terminate or yield – One thread: just execute that thread – More than one thread: now the scheduler has to make a • In a preemptive system, the scheduler can interrupt a resource allocation decision process that is running. Terminated • The scheduling algorithm determines how jobs New Ready Running are scheduled Waiting 3 4 Process States Scheduling Evaluation Metrics • There are many possible quantitative criteria for Terminated evaluating a scheduling algorithm: New Ready Running – CPU utilization: percentage of time the CPU is not idle – Throughput: completed processes per time unit Waiting – Turnaround time: submission to completion – Waiting time: time spent on the ready queue Processes
    [Show full text]
  • Gang Scheduling Java Applications with Tessellation
    Gang Scheduling Java Applications with Tessellation ∗ Benjamin Le Jefferson Lai Wenxuan Cai John Kubiatowicz {benjaminhoanle, jefflai2, wenxuancai}@berkeley.edu, [email protected] Department of Electrical Engineering and Computer Science University of California, Berkeley Berkeley, CA 94720-1776 ABSTRACT 1. INTRODUCTION Java Virtual Machines (JVM) are responsible for perform- The Java Virtual Machine (JVM) abstraction provides a ing a number of tasks in addition to program execution, consistent, contained execution environment for Java appli- including those related to garbage collection and adaptive cations and has played a significant role in Java's popularity optimization. Given the popularity of the Java platform and success in recent years. The JVM abstraction encapsu- and the rapid growth in use of cloud computing services, lates many tasks and responsibilities that make development cluster machines are expected to host a number of concur- faster and easier. For example, by handling the translation rently executing JVMs. Despite the virtual machine ab- of portable Java bytecode to machine code specific to the straction, interference can arise between JVMs and degrade lower level kernel and system architecture, the JVM allows performance. One approach to prevent such multi-tenancy developers to write an application once and run it in a num- issues is the utilization of resource-isolated containers. This ber of different environments. While there exist a number of paper profiles the performance of Java applications execut- JVM implementations, maintained both privately and pub- ing within the resource-isolated cells of Adaptive Resource- licly, in general, implementations adhere to a single JVM Centric Computing and the Tessellation OS.
    [Show full text]
  • CPU Scheduling
    CPU Scheduling CSC 256/456 - Operating Systems Fall 2014 TA: Mohammad Hedayati Agenda ●Scheduling Policy Criteria ●Scheduling Policy Options (on Uniprocessor) ●Multiprocessor scheduling considerations ●CPU Scheduling in Linux ●Real-time Scheduling Process States ●As a process executes, it changes state onew: The process is being created oready: The process is waiting to be assigned to a process orunning: Instructions are being executed owaiting: The process is waiting for some event to occur oterminated: The process has finished execution Queues of PCBs ● Ready Queue: set of all processes ready for execution. ● Device Queue: set of processes waiting for IO on that device. ● Processes migrate between the various queues. Scheduling ●Question: How the OS should decide which of the several processes to take of the queue? oWe will only worry about ready queue, but it is applicable to other queues as well. ●Scheduling Policy: which process is given the access to resources? CPU Scheduling ● Selects from among the processes/threads that are ready to execute, and allocates the CPU to it ● CPU scheduling may take place at: o hardware interrupt o software exception o system calls ● Non-Preemptive o scheduling only when the current process terminates or not able to run further (due to IO, or voluntarily calling sleep() or yield()) ● Preemptive o scheduling can occur at any opportunity possible Some Simplifying Assumptions ●For the first few algorithms: oUni-processor oOne thread per process oPrograms are independant ●Execution Model: Processes alternate between bursts of CPU and IO ●Goal: deal out CPU time in a way that some parameters are optimized. CPU Scheduling Criteria ● Minimize waiting time: amount of time the process is waiting in ready queue.
    [Show full text]
  • Lecture 4: Scheduling
    CS 318 Principles of Operating Systems Fall 2019 Lecture 4: Scheduling Prof. Ryan Huang Administrivia • Lab 0 - Due today - Submit in Blackboard • Lab 1 released - Due in two weeks - Lab overview session next week - If you still don’t have a group, let us know soon - GitHub classroom invitation link on Piazza post 9/12/19 CS 318 – Lecture 4 – Scheduling 2 Recap: Processes • The process is the OS abstraction for execution - own view of machine • Process components - address space, program counter, registers, open files, etc. - kernel data structure: Process Control Block (PCB) • Process states and APIs - state graph and queues - process creation, deletion, waiting • Multiple processes - overlapping I/O and CPU activities - context switch 9/12/19 CS 318 – Lecture 4 – Scheduling 3 Scheduling Overview • The scheduling problem: - Have � jobs ready to run - Have � ≥ 1 CPUs • Policy: which jobs should we assign to which CPU(s), for how long? - we’ll refer to schedulable entities as jobs – could be processes, threads, people, etc. • Mechanism: context switch, process state queues 9/12/19 CS 318 – Lecture 4 – Scheduling 4 Scheduling Overview 1. Goals of scheduling 2. Textbook scheduling 3. Priority scheduling 4. Advanced scheduling topics 9/12/19 CS 318 – Lecture 4 – Scheduling 5 When Do We Schedule CPU? ❸ ❹ ❷ ❶ ❸ • Scheduling decisions may take place when a process: ❶ Switches from running to waiting state ❷ Switches from running to ready state ❸ Switches from new/waiting to ready ❹ Exits • Non-preemptive schedules use ❶ & ❹ only • Preemptive
    [Show full text]
  • COS 318: Operating Systems CPU Scheduling
    COS 318: Operating Systems CPU Scheduling Jaswinder Pal Singh Computer Science Department Princeton University (http://www.cs.princeton.edu/courses/cos318/) Today’s Topics u CPU scheduling basics u CPU scheduling algorithms 2 When to Schedule? u Process/thread creation u Process/thread exit u Process thread blocks (on I/O, synchronization) u Interrupt (I/O, clock) 3 Preemptive and Non-Preemptive Scheduling Terminate Exited (call scheduler) Scheduler dispatch Running Block for resource (call scheduler) Yield, Interrupt (call scheduler) Ready Blocked Create Resource free, I/O completion interrupt (move to ready queue) 4 Scheduling Criteria u Assumptions l One process per user and one thread per process l Processes are independent u Goals for batch and interactive systems l Provide fairness l Everyone makes some progress; no one starves l Maximize CPU utilization • Not including idle process l Maximize throughput • Operations/second (min overhead, max resource utilization) l Minimize turnaround time • Batch jobs: time to execute (from submission to completion) l Shorten response time • Interactive jobs: time response (e.g. typing on a keyboard) l Proportionality • Meets user’s expectations Scheduling Criteria u Questions: l What are the goals for PCs versus servers? l Average response time vs. throughput l Average response time vs. fairness Problem Cases u Completely blind about job types l Little overlap between CPU and I/O u Optimization involves favoring jobs of type “A” over “B” l Lots of A’s? B’s starve u Interactive process trapped
    [Show full text]
  • Technical Report Nr. 745
    Technical Report Nr. 745 Systems Group, Department of Computer Science, ETH Zurich Gang scheduling isn’t worth it. yet by Simon Peter, Andrew Baumann, Zachary Anderson, Timothy Roscoe November 25, 2011 Gang scheduling isn’t worth it. yet Simon Peter,∗ Andrew Baumann,y Zachary Anderson,∗ Timothy Roscoe∗ ∗ Systems Group, Department of Computer Science, ETH Zurich y Microsoft Research Abstract programs typically have exclusive access to the machine The hardware trend toward higher core counts will likely (or partition thereof), and time slices are long (at least result in a dynamic, bursty and interactive mix of parallel 100ms) relative to context-switch time. applications in personal and server computing. We inves- In contrast, general-purpose OSes have not widely tigate whether gang scheduling can provide performance used gang schedulers, because parallel workloads that benefits for applications in this scenario. We present benefit from them have been rare, and techniques such a systematic study of the conditions under which gang as adaptive blocking synchronization are thought of as scheduling might be better than classical general-purpose sufficient to achieve near-optimal performance [20]. OS scheduling, and derive a set of necessary conditions In this paper we confirm that gang scheduling indeed on the workload. has few benefits for current general-purpose workloads, We find that these conditions are rarely met today, ex- except in a very limited set of cases. We characterize this cept in a small subset of workloads, for which we give space of workloads to show when gang scheduling can an example. However, we propose that this subset is deliver significant benefits over classical OS schedulers.
    [Show full text]