Abstract
The diversification of computing resources and the increasing complexity of resource demand from applications in terms of type, granularity, and quantity call for more efficiency in resource scheduling. To meet this challenge, this paper proposes a resource description model based on a quantized polygon. It explores the theoretical basis for the multilateral complementarity strategy (MCS), analyzes the basic mechanisms and application architecture of the MCS, and, finally, proposes the multi-dimensional quantized polygon (MQP) algorithm-a multi-dimensional resource scheduling algorithm based on multilateral complementarity (MC-MDRS algorithm). The experiments show that the multi-dimensional resource scheduling strategy based on the MQP, when implemented in an environment in which multi-dimensional (3 ≤ d ≤ 6) computing resources are provisioned, can effectively respond to the requests for various computing resources by granular services and facilitate the deployment of granular application services arriving in batches. The analysis of the experimental results indicates that the MQP algorithm outperforms other multi-dimensional resource scheduling algorithms by 2%-5% in node utilization.
Highlights
Since different types of resource requirements submitted to the system can be combined to form a complementary state, the algorithms based on multi-dimensional quantized polygon (MQP) strategy can effectively reduce resource fragmentation caused by unreasonable resource allocation process, and can legitimately integrate the complementary tasks effectively
A resource description model based on the quantitative polygon has been presented, the theoretical basis of multilateral complementarity strategy (MCS) has been explored, the basic mechanisms and application framework for implementing MCS have been analyzed, and a multi-dimensional resource scheduling algorithm based on multilateral complementarity-MQP has been proposed to meet the resource scheduling requirements during the deployment of granular application services in which the application services make multidimensional (3 ≤ d ≤ 6) resource requests to the GAC
Experiments were carried out to compare the performance of the MQP algorithm and other algorithms based on a greedy strategy and the vector dot product (FFDProd, FFDAvgSum, DP, L2, etc.), with the aim of verifying the effectiveness of the MQP algorithm in multi-dimensional resource scheduling
Summary
This chapter proposes a multi-dimensional resource scheduling strategy based on multilateral complementarity Implementing this strategy requires effective description and storage of three key types of information before resource scheduling: (1) basic information on the resource requests for various application services, and the ratios between the resources; (2) ratios between the occupied resources (quantized) at each server/computing node; (3) ratios between the remaining resources (quantized) at each server/computing node. The remainder of this paper is organized as follows: Section 2, briefly describes the issue of resource scheduling in data centers and gives a well-focused account of the relevant literature in this area It ends with a commentary on what is lacking in previous research, that existing methods cannot satisfy the need to ensure balanced resource utilization among multiple dimensions and that they lack a description of the dependency relationship between resource request, occupied resources and remaining resources.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
