Online Scheduling of Moldable Jobs with Deadline for HPC as a Service

• Main Authors: Wei Hsieh, 謝瑋
• Other Authors: Kuo Chan Huang
• Format: Others
• Language: en_US
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/40002120452761001338

碩士 === 國立臺中教育大學 === 資訊工程學系 === 102 === Traditionally, HPC systems usually deal with the so called best-effort jobs which do not have deadline and are scheduled in an as-quick-as-possible manner. Recently the concept of HPC as a Service (HPCaaS) was proposed to transform HPC facilities and applications into a more convenient and accessible service model. To achieve that goal, one inevitable work is to develop advanced job scheduling approaches for various types of HPCaaS jobs and take Quality of Service (QoS) into consideration. This thesis presents four scheduling approaches for online moldable jobs with deadline, which differ in how they arrange the first job in queue under currently available processors. Two important issues were explored when designing the four scheduling approaches. The first is concerned with waiting queue sequencing and three policies, EDF, LLF, and SJF, were evaluated. For the second part, we propose and evaluate three different processor allocation mechanisms: SNP, TLNP, and EST. The four scheduling approaches in combination with different waiting queue sequencing policies and allocation mechanisms are explored thoroughly towards the goal of HPC as a Service. In addition to traditional average turnaround time, the proposed approaches were evaluated and compared in terms of three other service-oriented performance metrics, including deadline miss rate, total profit, and lag of discard. We also evaluated the proposed scheduling approaches with a mix of best-effort jobs and deadline-constrained jobs, which is likely to be a common scenario in real HPCaaS systems.