Nested QoS: Providing Flexible SLAs in Shared Storage Systems

Electrical and Computer Engineering

Speaker: Hui Wang
Friday, February 3, 2012
10:00 AM to 11:00 AM


3110 Duncan Hall



The increasing popularity of storage and server consolidation introduces new challenges for resource management, capacity provisioning, and guaranteeing application performance. Typical Service Level Agreements (SLAs) provide performance guarantees in terms of throughput (IOPS) or response time limits (ms). The bursty nature of storage workloads [1] (where instantaneous arrival rates significantly exceed the average) implies a large gap between peak and average resource requirements in meeting response time bounds, leading to low overall server utilization and high cost. This situation is driving the development of elastic QoS models that allow clients greater flexibility in adopting SLAs tailored to their workload characteristics and performance requirements, while allowing the service provider opportunities to optimize provisioning and scheduling decisions. In this thesis we present a novel Nested QoS service model for multiplexing multiple concurrent bursty workloads on a shared storage server. Our solution employs two strategies together: systematically classifying requests to provide each workload with a graduated Quality of Service (QoS) and flexibly scheduling the classified portions of all the workloads. The results show that the Nested QoS service model provides (i) performance isolation and strong performance guarantees in spite of the badly behaved workload (2) a flexible and auditable elastic SLA definition (3) improved server utilization.