Title: Improving the Performance of Applications in Cloud Environments
Abstract:
The performance of applications in clouds are subjected to many problems. In this proposal, we mainly tackle two important problems in clouds – network and security. Network utilization has been identified as a key factor for the performance of data-parallel frameworks. First, to handle the network problem in data-parallel clusters, we propose a network-aware scheduler (NAS) that handles two main challenges associated with the shuffle phase for high performance: i) balancing cross-node network load, and ii) avoiding and reducing cross-rack network congestion. Second, to keep pace to the modern datacenter networks, we propose JobPacker, a job scheduler for data-parallel frameworks in hybrid electrical/optical datacenter network that aims to take full advantage of the optical circuit switch to improve the job performance. Third, to prevent applications from being impacted by the memory DoS attacks in clouds, we propose a comprehensive scheme to accurately detect the attacks for applications with different characteristics. The proposal provides an overview of the proposed solutions, proposed evaluation metrics, research plan and some preliminary results.
Committee members:
Haiying Shen (Advisor); Andrew Grimshaw (Chair); Mary Lou Soffa, David Evans, and Zongli Lin (Minor representative).