Ph.D. Dissertation Proposal Announcement

Title: Improving n-Tier Application Performance at High Resource Utilization: the Impact of Soft Resource Allocation and Management

Presented by
Qingyang Wang
School of Computer Science
College of Computing
Georgia Institute of Technology


Date: Thursday, May 3, 2012
Time: 10:00 AM - 12:00 PM
Where: KACB - Conference Room 1202


Committee:

• Professor Dr. Calton Pu, Advisor (School of Computer Science, College of Computing, Georgia Institute of Technology)

• Professor Dr. Ling Liu (School of Computer Science, College of Computing, Georgia Institute of Technology)
• Professor Dr. Leo Mark (School of Computer Science, College of Computing, Georgia Institute of Technology)
• Professor Dr. Shamkant B. Navathe (School of Computer Science, College of Computing, Georgia Institute of Technology)

Summary:
Simultaneously achieving good performance and high resource utilization is an important goal for production cloud environments. Through extensive measurements of n-tier web-facing application benchmarks, we show that overall system performance at high resource utilization is surprisingly sensitive to appropriate allocation and management of soft resources (server thread pool size or database connection pool size).
Inappropriate soft resource allocation and management can degrade overall application performance significantly. Concretely, both under-allocation and over-allocation of thread pool can lead to bottlenecks in other resources because of non-trivial dependencies among the software components and hardware components in the system. We have observed some non-obvious phenomena due to these correlated bottlenecks.
For instance, inappropriate soft resource allocation may lead to a challenging rapidly oscillating bottleneck phenomenon, limiting the throughput of the entire system while none of the hardware resources in the system is fully saturated. To address this challenge, we provide several techniques to carefully allocate and manage soft resources in the system through a systematic empirical approach based on fine-grained monitoring data in each tier of the system. our results show that soft resource allocation and management plays a central role for n-tier web-facing applications to achieve good performance at high resource utilization in cloud environments.