High Performance Computing (HPC) systems are subject to dynamical variations occurring in, e.g., jobs execution duration, I/O quantity, network consumption. Adapting to these unpredictable variations requires using autonomic management in an online feedback loop. The introduction of control theory methods allows for the design of well-founded autonomic managers. Choosing the relevant approach is daunting due to the variety of existing controllers. The criteria are of different natures, involving performance and efficiency, but also required expertise in control theory, and reusability or portability between sub-systems. Therefore, there is a need for comparative studies to assist designers choices. We consider the problem of resource harvesting in HPC systems, where scheduling often leaves resources idle. Our approach controls—through a feedback loop—the injection of small jobs in order to maximize the resources’ usage. The control problem is to manage the tradeoff between harvesting and performance, in a reusable manner. We study how reusability relates to the adaptivity and robustness properties in control. We illustrate our approach with the classic Proportional-Integral-Derivative (PID) control, its upgrade as adaptive control, and Model-Free Control (MFC). We target CiGri , a system harvesting idle resources in a computing grid. We perform experimental evaluation and compare performance and reusability. Tradeoffs are found on different criteria: While adaptive control is largely portable, its design complexity is significant for non-experts; PID control has good nominal performance, yet its portability is limited; MFC requires few competences to be used, but cannot provide strong guarantees.

We present a new approach for scheduling independent tasks on multiple CPUs and multiple GPUs. The tasks are assumed to be parallelizable on CPUs using the moldable model: the final number of cores allotted to a task can be decided and set by the scheduler. More precisely, we design an algorithm aiming at minimizing the makespan-the maximum completion time of all tasks-for this scheduling problem. The proposed algorithm combines a dual approximation scheme with a fast integer linear program (ILP). It determines both the partitioning of the tasks, i.e., whether a task should be mapped to CPUs or a GPU, and the number of CPUs allotted to a moldable task if mapped to the CPUs. A worst-case analysis shows that the algorithm has an approximation ratio of 3/2 + ε. Since the time complexity of the ILP-based algorithm could be non-polynomial, we also present a polynomial-time algorithm with an approximation ratio of 2 + ε. We complement the theoretical analysis of our two novel algorithms with a simulation study. In these simulations, we compare our algorithms to a modified version of the classical HEFT algorithm, which we adapted to handle moldable tasks. The simulation results show that our algorithm with the (3/2 + ε)-approximation ratio produces significantly shorter schedules than the modified HEFT for most of the instances. In addition, our results provide evidence that our ILP-based algorithm can solve larger problem instances in a reasonable amount of time.

Summary More and more computers use hybrid architectures combining multi‐core processors and hardware accelerators such as graphics processing units (GPUs). We present in this paper a new method for scheduling efficiently parallel applications with m CPUs and k GPUs, where each task of the application can be processed either on a core (CPU) or on a GPU. The objective is to minimize the maximum completion time (makespan). The corresponding scheduling problem is Non‐deterministic Polynomial (NP)‐time hard, Copyright © 2014 John Wiley & Sons, Ltd.