Integrated resource management for cluster-based Internet services

Abstract

Client request rates for Internet services tend to be bursty and thus it is important to maintain efficient resource utilization under a wide range of load conditions. Network service clients typically seek services interactively and maintaining reasonable response time is often imperative for such services. In addition, providing differentiated service qualities and resource allocation to multiple service classes can also be desirable at times. This paper presents an integrated resource management framework (part of Neptune system) that provides flexible service quality specification, efficient resource utilization, and service differentiation for cluster-based services. This framework introduces the metric of quality-aware service yield to combine the overall system efficiency and individual service response time in one flexible model. Resources are managed through a two-level request distribution and scheduling scheme. At the cluster level, a fully decentralized request distribution architecture is employed to achieve high scalability and availability. Inside each service node, an adaptive scheduling policy maintains efficient resource utilization under a wide range of load conditions. Our trace-driven evaluations demonstrate the performance, scalability, and service differentiation achieved by the proposed techniques.

Pricing-based strategies for autonomic control of web servers for time-varying request arrivals

“…It shows that the efficiency benefits from controlled overbooking of resources can be dramatic. Similar ideas for an integrated resource management framework have been proposed in (Shen et al, 2002). It introduces the metric quality-aware service yield to combine the overall system efficiency and individual service response time in one flexible model.…”

Resource allocation optimization for quantitative service differentiation on server clusters

“…In [23], Shen et al proposed and designed a sound integrated resource management framework that provides flexible service quality specification, efficient resource utilization, and service differentiation for cluster-based Internet services. The work introduced an interesting metric, quality-aware service yield, to combine the overall system efficiency and individual service response time in one model.…”

“…7: prohibit the process on port i from listening new requests; // this process will become idle and killed by Apache. 8: end for 9: end for In each process allocation period, a multiplier m is used to keep the ratio of the number of active processes of process pools to the normalized value specified by the partial node partition (23). At line 3, µ * i is the normalized integer value of the number of processes allocated to the process pool i.…”

Quality-of-service differentiation on the Internet: A taxonomy

“…There are also other efforts on OS support for QoS differentiation on cluster of servers. For example, the work in Shen et al (2002) proposed an integrated two-level resource management framework that supports QoS differentiation in cluster-based Internet services.…”

“…Instead, class-based QoS differentiation is a relatively new concept and many studies have been devoted to this area in recent years. Examples include packet scheduling and dropping for DiffServ in the network core Ramanathan (1999, 2000;Liao and Campbell, 2004;Wang et al, 2004), multicasting in DiffServ domains (Li and Mohapatra, 2003), controltheoretical approaches for relative QoS differentiation (Abdelzaher et al, 2002;Lu et al, 2005), QoS differentiation in Web servers (Chen and Mohapatra, 2002;Zhang et al, 2003), resource management for QoS differentiation in cluster-based Internet servers and Grids (Shen et al, 2002;Yeow and Tham, 2004;Zhu et al, 2001), OS and middleware support for QoS isolation (Shenoy et al, 2002;Sundaram et al, 2000). Though many significant research results exist, however, there is no survey paper in the QoS differentiation context.…”