On-demand, service-oriented cloud computing infrastructures continue to increase in popularity with organizations. Three observations motivate us to investigate running high-throughput, data-intensive tasks as background workloads on these cloud infrastructures. First, the rapid growth in hardware parallelism leaves more residue resources to be exploited. Second, the ``incremental power usage'' of piggybacking a secondary background workload onto the foreground workload to utilize those residue resources is relatively low. Third, the advances in GPGPU (General-Purpose GPU) processing enable a novel coupling of concurrent workloads.
This project will explore a new computing model of offering cloud services on active nodes that are serving on-demand utility computing users. We plan to (1) assess the efficacy of resource sharing between foreground and background workloads and investigate the relationship between their resource usage patterns and the benefit and cost of their mixed execution; (2) develop scheduling and load management middleware that performs dynamic background workload distribution considering the energy-performance tradeoff; and (3) exploit the use of GPGPUs for cloud services on active nodes that are running foreground workloads mainly on the CPUs.
Our research will explore a revolutionary change in the use of cloud computing and may influence their hosting organizations' future resource configuration and planning to create greener clouds. The research will be closely integrated with education-oriented cloud platforms at NCSU. The PIs will also leverage their established services and connections to increase the participation of women and minority students and to promote students' interactions with industry partners.
With the rapid emergence of service-oriented cloud computing infrastructures, both public (e.g., Amazon EC2 and Microsoft Azure) and private (e.g., Cloudera and Rackspace), the efficient use of such cloud resources with respect to performance and power consumption is imperative for cloud computing to become a cost-effective, next-generation solution for the masses. Unfortunately, current cloud infrastructures typically operate at only 20%-50% utilization and arguably deliver mediocre performance while consuming more power and energy than necessary. Such waste translates into higher operational cost, which in turn is passed on to the consumer. This project sought to address the above shortcomings by exploring a new operational model for cloud computing, one where background jobs are scheduled (or "piggy-backed") on compute nodes with existing foreground jobs so as to increase resource utilization from the typical 20-50% to a more aggressive 70-90%. The techniques that were researched and developed in this project deliver cost-effective and non-intrusive background computing by considering the behavioral patterns of both the foreground and background workloads. This approach differs from existing power and energy management studies, in the sense that we pack additional work on a compute node at a low incremental cost with respect to power and energy, while existing approaches strive to slow down or shut down resources when they do not contribute to higher application productivity. The end result of the project is a myriad of technologies that improve performance and reduce power and energy consumption, and in turn, improve the energy efficiency of cloud computing environments. For example, MOON, short for MapReduce On Opportunistic eNvironments, leverages the MapReduce framework that Google uses for its seach engine and adapts it to cleverly harvest the spare cycles of idle desktop computers, thus delivering a highly cost-effective and energy-efficient private cloud for the masses.
