Greedy Allocation Scheme Research Articles

Parallel execution of schedules with random dependency graph

We consider schedules of unit-duration tasks with random dependencies. Formally the schedules are obtained by imposing a random partial order of vertices in an Erdős-Rényi graph. For such schedules, we provide asymptotic formulas for the parallel execution time and for the required number of processors in a greedy allocation scheme as a function of connection probability. We also derive asymptotic bounds for the number of required processors in an optimal allocation scheme. We test our results for small schedule sizes using simulation and conclude that the convergence to asymptotic results is achieved relatively fast in practice. In our simulation experiment, we define and compare the efficiency of Mixed Integer Programming and Constraint Programming approaches to find the exact solution of the optimal allocation scheme and conclude that Constraint Programming is more efficient in our setting. In the last part of the paper, we provide preliminary results of similar analysis for random schedules generated from arbitrary graphs.

A Reinforcement Learning-Based Resource Allocation Scheme for Cloud Robotics

In recent years, robotic systems combined with cloud computing capability have become an emerging topic of discussion in academic fields. The concept of cloud robotics allows the system to offload computing-intensive tasks from the robots to the cloud. An appropriate resource allocation scheme is necessary for the cloud computing service platform to efficiently allocate its computing resources, when the robots send requests asking for computing service. This paper proposes a resource allocation scheme based on reinforcement learning (RL), which can make the cloud to decide whether a request should be accepted and how many resources are supposed to be allocated. The scheme realizes an autonomous management of computing resources through online learning, reduces human participation in scheme planning, and improves the overall utility of the system in the long run. Numerical results demonstrate that the proposed RL-based computing resource allocation scheme has better performances than the greedy allocation scheme.

Low Complexity Resource Allocation Algorithm by Multiple Attribute Weighing and User Ranking for OFDMA Systems

We propose an effective subcarrier allocation scheme for multiuser orthogonal frequency division multiple access (OFDMA) system in the downlink transmission with low computational complexity. In the proposed scheme, by taking multiple attributes of a user's channel, such as carrier gain decrease rate and variation from the mean channel gain of the system, to determine a rank for the user, subcarriers are then allocated depending on the individual user's rank. Different channel characteristics are used to better understand a user's need for subcarriers and hence determine a priority for the user. We also adopt an attribute weighing scheme to enhance the performance of the proposed scheme. The scheme is computationally efficient, since it avoids using iterations for the algorithm convergence and also common water-filling calculations that become more complex with increasing system parameters. Low complexity is achieved by allocating subcarriers to users depending on their determined rank. Our proposed scheme is simulated in comparison with other mathematically efficient subcarrier allocation schemes as well as with a conventional greedy allocation scheme. It is shown that the proposed method demonstrates competitive results with the simulated schemes.

Optimality Aspects of Greedy Schemes in Parallel Processing of Random Graph-Structured Jobs

Parallel processing systems with jobs structured as random graphs, where the nodes correspond to executable tasks and the directed edges to precedence constraints, are studied from a queueing theoretic point of view under general stationarity assumptions on the job flows. Jobs need to have their tasks processed non-preemptively by a set of uniform processors. Simple, natural greedy schemes of allocating processors to tasks are shown to asymptotically minimize the long-term average execution time per job. The stability condition for this queueing system is specified, and greedy allocation schemes are shown to stabilize the system under the maximum possible job arrival rate. Some recurrence properties of the system state are also established.