As cloud computing emerges, software system has changed from standalone applications into service-based systems based on the Software as a Service (SaaS) model. Recently the concept of HPC as a Service (HPCaaS) was proposed to transform High-Performance Computing (HPC) facilities and applications into a more convenient and accessible service model. To achieve that goal, one inevitable issue of developing such kind of software systems is dealing with parallel job scheduling, taking Quality of Service (QoS) into consideration. This paper presents two new scheduling approaches for online moldable jobs with deadline, which differ in how they backfill the jobs in queue. In addition to traditional average turnaround time, the proposed approaches were evaluated and compared to previous approaches in terms of two other service-oriented performance metrics, including deadline miss rate and total profit, with a series of simulation experiments. The experimental results indicate that the new approaches outperform previous methods significantly in different performance metrics.