Coordination mechanisms for selfish scheduling

In machine scheduling, a set of jobs must be scheduled on a set of machines so as to minimize some global objective function, such as the makespan, which we consider in this paper. In practice, jobs are often controlled by independent, selfishly acting agents, which each select a machine for processing that minimizes the (expected) completion time. This scenario can be formalized as a game in which the players are job owners, the strategies are machines, and a player’s disutility is the completion time of its jobs in the corresponding schedule. The equilibria of these games may result in larger-than-optimal overall makespan. The price of anarchy is the ratio of the worst-case equilibrium makespan to the optimal makespan. In this paper, we design and analyze scheduling policies, or coordination mechanisms, for machines which aim to minimize the price of anarchy of the corresponding game. We study coordination mechanisms for four classes of multiprocessor machine scheduling problems and derive upper and lower bounds on the price of anarchy of these mechanisms. For several of the proposed mechanisms, we also prove that the system converges to a pure-strategy Nash equilibrium in a linear number of rounds. Finally, we note that our results are applicable to several practical problems arising in communication networks.     

应用团划分方法改进多处理机任务近似调度

研究多处理机任务调度模型P_m|fix,p_j=1|C_max,即在m个处理机系统中调度n个时间长度都为1的多处理机任务,每个任务指派到所需一组处理机上不可剥夺地执行。这类问题在网络并行计算、多播系统及工程规划等领域都有广泛的应用,但早已被证明为NP难问题,而且也不存在常数近似算法。基于团划分方法构造了该问题的多项式时间近似算法,通过模拟实验进行了验证,和最大宽度优先（LWF）算法相比,该算法花费时间较长,近似比性能要好。     

Programming support and scheduling for communicating parallel tasks

Task-based programming models are beneficial for the development of parallel programs for several reasons. They provide a decoupling of the specification of parallelism from the scheduling and mapping to execution resources of a specific hardware platform, thus allowing a flexible and individual mapping. For platforms with a distributed address space, the use of parallel tasks, instead of sequential tasks, adds the additional advantage of a structuring of the program into communication domains that can help to reduce the overall communication overhead.     

移动Agent任务调度遗传算法

提出了一种可在全部可行解空间寻求最优解的混沌遗传算法。 算法定义了问题模型及染色体表示方法; 生成了考虑通信代价的贪婪算法初始种群; 设计了一种新的交叉机制来确保杂交后生成的解依然为合法解; 使用了启发变异算子; 采用混沌优化技术动态控制交叉、变异操作; 还使用了父代若干精英个体参与选择策略。 最后给出了3种通信代价、2种任务节点及3种主机节点共18个任务图的仿真结果。     

On scheduling tasks with exponential service times and in-tree precedence constraints

Summary In this paper we extend the work of Chandy and Reynolds in which they considered the problem of scheduling tasks on two identical processors. The processing times of the tasks are independent, identically distributed exponential random variables. The tasks are subject to in-tree precedence constraints. Chandy and Reynolds have shown that the expected value of the makespan is minimized if and only if the scheduling policy is Highest-Level-First (HLF). We extend their result by showing that a policy maximizes the probability that all tasks finish by some time t0 if and only if the policy is HLF. Additionally, we show that a policy maximizes the probability that the sum of the finishing times of all the tasks is less than some value s0 if and only if the policy is HLF.This work has been partially supported by a Lady Davis Fellowship at the Technion, Haifa Israel; by the Research Institute for Advanced Computer Science at the NASA Ames Research Center, Moffett Field CA; and by NSF Grant MCS 80-04257.     

On approximation of the bulk synchronous task scheduling problem

The bulk synchronous task scheduling problem (BSSPO) is known as an effective task scheduling problem for distributed memory machines. We present a proof of NP-completeness of the decision counterpart of BSSPO, even in the case of makespan of at most five. This implies nonapproximability of BSSPO, meaning that there is no approximation algorithm with performance guarantee smaller than 6/5 unless P = NP. We also give an approximation algorithm with a performance guarantee of two for BSSPO in several restricted cases.     

Preference-oriented fixed-priority scheduling for periodic real-time tasks

Traditionally, real-time scheduling algorithms prioritize tasks solely based on their timing parameters and cannot effectively handle tasks that have different execution preferences. In this paper, for a set of periodic real-time tasks running on a single processor, where some tasks are preferably executed as soon as possible (ASAP) and others as late as possible (ALAP), we investigate Preference-Oriented Fixed-Priority (POFP) scheduling techniques. First, based on Audsley’s Optimal Priority Assignment (OPA), we study a Preference Priority Assignment (PPA) scheme that attempts to assign ALAP (ASAP) tasks lower (higher) priorities, whenever possible. Then, by considering the non-work-conserving strategy, we exploit the promotion times of ALAP tasks and devise an online dual-queue based POFP scheduling algorithm. Basically, with the objective of fulfilling the execution preferences of all tasks, the POFP scheduler retains ALAP tasks in the delay queue until their promotion times while putting ASAP tasks into the ready queue right after their arrivals. In addition, to further expedite (delay) the executions of ASAP (ALAP) tasks using system slack, runtime techniques based on dummy and wrapper tasks are investigated. The proposed schemes are evaluated through extensive simulations. The results show that, compared to the classical fixed-priority Rate Monotonic Scheduling (RMS) algorithm, the proposed priority assignment scheme and POFP scheduler can achieve significant improvement in terms of fulfilling the execution preferences of both ASAP and ALAP tasks, which can be further enhanced at runtime with the wrapper-task based slack management technique.     

A greedy approximation algorithm for minimum-gap scheduling

We consider scheduling of unit-length jobs with release times and deadlines, where the objective is to minimize the number of gaps in the schedule. Polynomial-time algorithms for this problem are known, yet they are rather inefficient, with the best algorithm running in time \(O(n^4)\) and requiring \(O(n^3)\) memory. We present a greedy algorithm that approximates the optimum solution within a factor of 2 and show that our analysis is tight. Our algorithm runs in time \(O(n^2 \log n)\) and needs only O(n) memory. In fact, the running time is \(O(n (g^*+1)\log n)\), where \(g^*\) is the minimum number of gaps.     

Multi-queue scheduling of two tasks

Summary Scheduling in the central server queueing model, consisting of a CPU server and m I/O servers, is considered for the case of two customers. Optimal (maximal CPU utilization) CPU and I/O schedules are obtained. The best CPU schedule depends on the I/O schedule in effect; and is either Longest or Shortest-Expected-Remaining-Processing-Time-First. However, for certain I/O schedules the CPU schedule is immaterial. The best I/O schedule is always to process the longer CPU customer first.     

Quasi-static scheduling of communicating tasks

Good scheduling policies for distributed embedded applications are required for meeting hard real time constraints and for optimizing the use of computational resources. We study the quasi-static scheduling problem in which (uncontrollable) control flow branchings can influence scheduling decisions at run time. Our abstracted distributed task model consists of a network of sequential processes that communicate via point-to-point buffers. In each round, the task gets activated by a request from the environment. When the task has finished computing the required responses, it reaches a pre-determined configuration and is ready to receive a new request from the environment. For such systems, we prove that determining the existence of a scheduling policy that guarantees upper bounds on buffer capacities is undecidable. However, we show that the problem is decidable for the important subclass of “data-branching” systems in which control flow branchings are exclusively due to data-dependent internal choices made by the sequential components. This decidability result exploits ideas derived from the Karp and Miller coverability tree for Petri nets as well as the existential boundedness notion of languages of message sequence charts.     

支持动态可重构硬件透明编程的预配置调度

面向微处理器和可编程器件加速器的混合异构多核体系结构的可重构计算环境,采用程序员熟悉的函数描述格式,在运行时根据软硬件划分的结果,动态实现到软件函数实体代码或者硬件函数实现电路的连接。为降低重配置开销,提高系统性能,统计了各个硬件函数的调用次数和次序,并结合其运行时间和硬件面积等信息,设计了一种预配置算法,尽量使配置和计算能够重叠处理,从而缩短系统的整体运行时间,获得更大性能加速。     

大规模短时间任务的低延迟集群调度框架

大规模数据分析环境中,经常存在一些持续时间较短、并行度较大的任务。如何调度这些低延迟要求的并发作业是目前研究的一个热点。现有的一些集群资源管理框架中,集中式调度器由于主节点的瓶颈无法达到低延迟的要求,而一些分布式调度器虽然达成了低延迟的任务调度,但在最优资源分配以及资源分配冲突方面存在一定的不足。从大规模实时作业的需求出发,设计和实现了一个分布式的集群资源调度框架,以满足大规模数据处理的低延迟要求。首先提出了两阶段调度框架以及优化后的两阶段多路调度框架;然后针对两阶段多路调度过程中存在的一些资源冲突问题,提出了基于负载平衡的任务转移机制,从而解决了各个计算节点的负载不平衡问题;最后使用实际负载以及一个模拟调度器对大规模集群中的任务调度框架进行了模拟和验证。对于实际负载,所提框架的调度延迟控制在理想调度的12%以内;在模拟环境下,该框架与集中式调度器相比在短时间任务的延迟上能够减少40%以上。     

Using job-shop scheduling tasks for evaluating collocated collaboration

Researchers have begun to explore tools that allow multiple users to collaborate across multiple devices in collocated environments. These tools often allow users to simultaneously place and interact with information on shared displays. Unfortunately, there is a lack of experimental tasks to evaluate the effectiveness of these tools for information coordination in such scenarios. In this article, we introduce job-shop scheduling as a task that could be used to evaluate systems and interactions within computer-supported collaboration environments. We describe properties that make the task useful, as well as evaluation measures that may be used. We also present two experiments as case studies to illustrate the breadth of scenarios in which this task may be applied. The first experiment shows the differences when users interact with different communicative gesturing schemes, while the second demonstrates the benefits of shared visual information on large displays. We close by discussing the general applicability of the tasks.     

An optimal scheduling algorithm for preemptable real-time tasks

An optimal scheduling algorithm is presented for real-time tasks with arbitrary ready times and deadlines in single processor systems. The time complexity of the algorithm is O(n log n), which improves the best previous result of O(n²). Furthermore, the lower bound of the worst-case time complexity of the problem is shown to be of O(n log n) and therefore the time complexity of the presented algorithm is shown to be lower bound.     

复合并行机排序问题的启发式算法研究

为有效解决复合并行机排序的极小化最大完成时间问题,提出了分支定界算法和改进的启发式动态规划算法。利用分支定界算法的3个工具:分支模型、边界和优先规则,构建出分支搜索树。按优先规则进行定界搜索,从而减小了问题求解规模。将原始作业转换为虚拟作业,根据Johnson法则,求解出原问题的最优排序。改进的动态规划算法复杂度分析和计算实验表明,这两个算法可靠性高并且可以解决实际问题。     

On the complexity of fixed-priority scheduling of periodic, real-time tasks

We consider the complexity of determining whether a set of periodic, real-time tasks can be scheduled on m 1 identical processors with respect to fixed-priority scheduling. It is shown that the problem is NP-hard in all but one special case. The complexity of optimal fixed-priority scheduling algorithm is also discussed.     

Allocation and scheduling of precedence-related periodic tasks

This paper discusses a static algorithm for allocating and scheduling components of periodic tasks across sites in distributed systems. Besides dealing with the periodicity constraints, (which have been the sole concern of many previous algorithms), this algorithm handles precedence, communication, as well as replication requirements of subtasks of the tasks. The algorithm determines the allocation of subtasks of periodic tasks to sites, the scheduled start times of subtasks allocated to a site, and the schedule for communication along the communication channel(s). Simulation results show that the heuristics and search techniques incorporated in the algorithm are very effective