Gabriele Capannini

Comparison of Multi-Criteria Scheduling Techniques

We propose a novel schedule-based approach for scheduling a continuous stream of batch jobs on the machines of a computational Grid. Our new solutions represented by dispatching rule Earliest Gap-Earliest Deadline First (EG-EDF) and Tabu search are based on the idea of filling gaps in the existing schedule. EG-EDF rule is able to build the schedule for all jobs incrementally by applying technique which fills earliest existing gaps in the schedule with newly arriving jobs. If no gap for a coming job is available EG-EDF rule uses Earliest Deadline First (EDF) strategy for including new job into the existing schedule. Such schedule is then optimized using the Tabu search algorithm moving jobs into earliest gaps again. Scheduling choices are taken to meet the Quality of Service (QoS) requested by the submitted jobs, and to optimize the usage of hardware resources. Proposed solution is compared with FCFS, EASY backfilling, and Flexible backfilling. Experiments shows that EG-EDF rule is able to compute good assignments, often with shorter algorithm runtime w.r.t. the other queue-based algorithms. Further Tabu search optimization results in higher QoS and machine usage.

Backfilling Strategies for Scheduling Streams of Jobs On Computational Farms

This paper presents a set of strategies for scheduling a stream of batch jobs on the machines of a heterogeneous computational farm. Our proposal is based on a flexible backfilling, which schedules jobs according to a priority assigned to each job submitted for execution. Priority values are computed as a result of a set of heuristics whose main goal is to improve resources utilization and to meet the job QoS requirements. The heuristics consider job deadlines, estimated execution time and aging of the jobs in the scheduling queue. Furthermore, the set of software licenses required by a job is also considered. The different proposals have been compared through simulations. Performance figures show the applicability of our approach.

A job scheduling framework for large computing farms

In this paper, we propose a new method, called Convergent Scheduling, for scheduling a continuous stream of batch jobs on the machines of large-scale computing farms. This method exploits a set of heuristics that guide the scheduler in making decisions. Each heuristics manages a specific problem constraint, and contributes to carry out a value that measures the degree of matching between a job and a machine. Scheduling choices are taken to meet the QoS requested by the submitted jobs, and optimizing the usage of hardware and software resources. We compared it with some of the most common job scheduling algorithms, i.e. Backfilling, and Earliest Deadline First. Convergent Scheduling is able to compute good assignments, while being a simple and modular algorithm.

A parallel code for time independent quantum reactive scattering on CPU-GPU platforms

The innovative architecture of GPUs has been exploited to the end of implementing an efficient version of the time independent quantum reactive scattering ABC code. The intensive usage of the code as a computational engine for several molecular calculations and crossed beams experiment simulations has prompted a detailed analysis of the utilization of the innovative features of the GPU architecture. ABC has shown to rely on a heavy usage of blocks of recursive sequences of linear algebra matrix operations whose performances vary significantly with the input and the section of the code. This has requested the evaluation of the suitability of different implementation strategies for the various parts of ABC. The outcomes of the related test runs are discussed in the paper.

K-Model: A New Computational Model for Stream Processors

We introduce K-model, a computational model to evaluate the algorithms designed for graphic processors, and other architectures adhering to the stream programming model. We address the lack of a formal complexity model that properly accounts for memory contention, address coalescing in memory accesses, or the serial control of instruction flows. We study the impact of K-model rules on algorithm design. We devise a coalesced and low contention data access technique for Batchers networks, and we evaluate the effectiveness of this technique within our K-model. To evaluate the benefits in using K-model in evaluating solutions for streaming architectures, we compare the complexity of a sorting network built using our technique, and quick sort. Although in theory quick sort is more efficient than bitonic sort, empirically, our bitonic sorting network has been shown to be faster than the state-of-the-art implementation of quick sort on graphics processing units (GPUs). We use our K-model to prove that this observation should generally hold. As a side result, our technique to perform a Batchers network on GPUs improves the performance of one the fastest comparison-based solution for integers sorting.

A Multi-criteria Job Scheduling Framework for Large Computing Farms

In this paper, we propose a multi-criteria job scheduler for scheduling a continuous stream of batch jobs on largescale computing farms, called Convergent Scheduling 2.0 (CS 2.0), which is an enhancement of the scheduler described in [5]. CS 2.0 exploits a set of heuristics that drive the scheduler in taking decisions. Each heuristics manages a specific constraint, and contributes to compute the measurement of the matching degree between a job and a machine. Scheduling choices are taken both to meet the QoS requested by the submitted jobs and to optimize the exploitation of hardware and software resources. In order to validate CS 2.0, we compared it versus two common job scheduling algorithms: Easy and Flexible backfilling. CS 2.0 demonstrated to be able to compute good assignments that allow a better exploitation of resources with respect to the other algorithms.

A multi-level scheduler for batch jobs on grids

This paper proposes a two-level scheduler for dynamically scheduling a continuous stream of sequential and multi-threaded batch jobs on grids, made up of interconnected clusters of heterogeneous single-processor and/or symmetric multiprocessor machines. The scheduler aims to schedule arriving jobs respecting their computational and deadline requirements, and optimizing the hardware and software resource usage. At the top of the hierarchy a lightweight meta-scheduler (MS) classifies incoming jobs according to their requirements, and schedules them among the underlying resources balancing the workload. At cluster level a Flexible Backfilling algorithm carries out the job machine associations by exploiting dynamic information about the environment. Scheduling decisions at both levels are based on job priorities computed by using different sets of heuristics. The different proposals have been compared through simulations. Performance figures show the feasibility of our approach.

Designing Efficient Parallel Prefix Sum Algorithms for GPUs

This paper presents a novel and efficient method to compute one of the simplest and most useful building block for parallel algorithms: the parallel prefix sum operation. Besides its practical relevance, the problem achieves further interest in parallel-computation theory. We firstly describe step-by-step how parallel prefix sum is performed in parallel on GPUs. Next we propose a more efficient technique properly developed for modern graphics processors and alike processors. Our technique is able to perform the computation in such a way that minimizes both memory conflicts and memory usage. Finally we evaluate theoretically and empirically all the considered solutions in terms of efficiency, space complexity, and computational time. In order to properly conduct the theoretical analysis we used a novel computational model proposed by us in a previous work: K-model. Concerning the experiments, the results show that the proposed solution obtains better performance than the existing ones.

A two-level scheduler to dynamically schedule a stream of batch jobs in large-scale grids

This paper describes the study conducted to design and evaluate a two-level on-line scheduler to dynamically schedule a stream of sequential and multi-threaded batch jobs on large scale grids, made up of interconnected clusters of heterogeneous machines. The scheduler aims to schedule arriving jobs respecting their computational and deadline requirements, and optimizing the utilization of hardware resources as well as software resources.