H. Zhu, H. Tang, and T. Yang, "Demand-Driven Service Differentiation in Cluster-Based Network Servers," Proc. IEEE Infocom, Apr. 2001.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....different classes of service have latencies within pre specified targets [9] Unlike our paper that focus on static and dynamic requests, they consider Web sites providing static content only. Two dynamic resource partitioning algorithms for static and dynamic Web requests have been proposed in [5, 15]. Their experiments demonstrate that dynamic server partitioning always outperforms static server partitioning. An interesting work from Aron et al. has extended the resource principal abstraction to multi node Web servers [2] Several companies commercialize as their most recent products ....

....Nortel Networks Alteon WebOS, F5 s BIG IP, Resonate s Central Dispatch) These switches provide only very simple service differentiation mechanisms which aim to statically partition server nodes and assign different classes of requests to different server subsets. Various results in literature [5, 15] demonstrate that static partitioning policies cannot adapt to fluctuating arrival rates and servers load conditions. Moreover, it may lead to waste of resources when some partitions are not fully utilized while others might be overloaded. The rest of this paper is organized as follows. Section 2 ....

H. Zhu, H. Tang, and T. Yang. Demand-driven service differentiation in cluster-based network servers. In Proc. of IEEE Infocom 2001.

....to maintaining lifetime targets for its premium objects, it would need to dynamically increase to maintain S1 at its target level. V. RELATED WORK Service differentiation has been applied to the networking domain [15 16] and more recently, in a flurry of activity, to the web server domain [10,17 25]. In the latter case, various mechanisms and policies are proposed for delivering end toend QoS to the origin servers to provide preferential service to preferred clients or services. Cache partitioning has been studied in a variety of domains. The 2Q algorithm [26] which is built upon the LRU K ....

H. Zhu, H. Tang, and T. Yang. Demand-driven service differentiation in cluster-based network servers. In IEEE Infocom, Anchorage, Alaska, Apr. 2001.

....the services. A major goal of the work is to maximize the number of unused servers so that they can be powered down to reduce energy consumption. Other mechanisms for shared server utilities include cluster reserves for multiple service performance isolation [4] and QoS differentiation algorithms [20]. Examples of commercial shared server utility computing products that partition server resources in clusters of shared servers include [12] 19] which have resource management systems that aim to satisfy SLAs while better supporting increased resource utilization. In contrast, a full server ....

H. Zhu, H. Tang, and T. Yang. Demand-driven service differentiation for cluster-based network servers. In Proceedings of IEEE INFOCOM'2001.

.... architecture integrated with admission control and performance isolation mechanisms has been proposed in [7] Some recent research efforts focus on how providing QoS support through the Web switch, by taking into account request information at layer 4 switches [23] as well as at layer 7 switches [30, 103], where a detailed content aware information allows to achieve performance isolation in Web clusters at a server level granularity. In particular, resource utilization can be improved by dynamically adjusting server partitions based on fluctuating requests arrival rates and servers load conditions ....

....where a detailed content aware information allows to achieve performance isolation in Web clusters at a server level granularity. In particular, resource utilization can be improved by dynamically adjusting server partitions based on fluctuating requests arrival rates and servers load conditions [22, 103]. While layer 4 Web cluster architectures may be considered a solved problem, the area of content aware architectures needs further research. Dispatching algorithms that combine effectively client and server information, and adaptive policies are not yet fully explored. Some companies ....

H. Zhu, H. Tang, and T. Yang. Demand-driven service differentiation in cluster-based network servers. In Proc. of IEEE Infocorn 2001.

....Monitor. We further investigate the relationship between request queue length (the waiting requests and those being served) and the request processing time under heavy server load. Since the processing time for individual URL s varies, for fair comparison, we adopt the stretch factor from [23], which refers to the quotient between the current processing time and the processing time of the same URL under normal load. The stretch factor reflects We refrain from mentioning the name of the retailer honoring a nondisclosure agreement. 2 3 4 5 6 7 8 9 10 0 5 10 15 20 25 ....

H. Zhu, H. Tang, T. Yang, "Demand-Driven Service Differentiation in Cluster-Based Network Servers," In Proceedings of the IEEE INFOCOM 2001.

....time for individual URL s varies, for fair comparison, 1 We refrain from mentioning the name of the retailer honoring a nondisclosure agreement. 1 2 3 4 5 6 7 8 9 10 0 5 10 15 20 25 Time(hour) Fig. 2. Traffic histogram of the server trace for a day. we adopt the stretch factor from [23], which refers to the quotient between the current processing time and the processing time of the same URL under normal load. The stretch factor reflects the current server load. Figure 3 depicts the queue length and stretch factor during the 20:00 21:00 period. It is observed that the two curves ....

H. Zhu, H. Tang, T. Yang, "Demand-Driven Service Differentiation in Cluster-Based Network Servers," In Proceedings of the IEEE INFOCOM

....prefetching [16] 27] Examples of approaches that aim to improve resource utilization include the elimination of unnecessary memory transfers between the various layers in the system (user space, kernel space, network buffers, etc. 29] and avoiding overloading through proper admission control [33]. Scalable Content Delivery Engines: To put our work in context, we compare it to recent projects that parallel some aspects of our two pronged approach to the construction of popular content delivery servers. Recall that our strategy relies on (1) an (almost) perfect sharing of memory between ....

H. Zhu, H. Tang, and T. Yang. Demand-driven service differentiation for cluster-based network servers. In Proceedings of IEEE INFOCOM, 2001.

....prefetching [16] 27] Examples of approaches that aim to improve resource utilization include the elimination of unnecessary memory transfers between the various layers in the system (user space, kernel space, network buffers, etc. 29] and avoiding overloading through proper admission control [33]. Scalable Content Delivery Engines: To put our work in context, we compare it to recent projects that parallel some aspects of our two pronged approach to the construction of popular content delivery servers. Recall that our strategy relies on (1) an (almost) perfect sharing of memory between 1 ....

H. Zhu, H. Tang, and T. Yang. Demand-driven service differentiation for cluster-based network servers. In Proceedings of IEEE INFOCOM, 2001.

....during some interval, which is easily mea sured. If # i #r iand # i ## target then the customer is underprovisioned and the center pays a penalty. For example, the penalty i function could increase with the degree of the shortfall r i =# i ,or with the resulting queuing delays or stretch factor [47]. The center must balance the revenue increase from overbooking against the risk of incurring a penalty. The penalty may be a partial refund of the reservation price, or if the penalty exceeds the bid then U i may take on negative values reflecting the customer s negative utility of reserving ....

....front ends or switches to virtualize network services (e.g. 32, 36, 7, 4] for scalability, availability, or both. Muse uses reconfigurable redirecting switches as a mechanism to support adaptive resource provisioning in network server pools; related projects considering this approach are DDSD [47] and Oceano [5] The switch features needed are similar to the support for virtual LANs, request load balancing, and server failover found on commercial switches for server traffic management, which are now widely used in large Internet server sites. Service hosting. Our focus on hosting services ....

Huican Zhu, Hong Tang, and Tao Yang. Demand-driven Service Differentiation in Cluster-Based Network Servers. In Proceedings of IEEE Infocom 2001, April 2001.

....[17] 35] 19] Examples of approaches that aim to improve resource utilization include the elimination of unnecessary memory transfers between the various layers in the system (user space, kernel space, network buffers, etc. 37] and avoiding overloading through proper admission control [41]. Scalable Content Delivery Engines: To put our work in context, we compare it to recent projects that parallel some aspects of our two pronged approach to the construction of popular content delivery servers. Recall that our strategy 2 In this paper, we use the term Internet servers to refer to ....

H. Zhu, H. Tang, and T. Yang. Demand-driven service differentiation for cluster-based network servers. In Proceedings of IEEE INFOCOM, 2001.

No context found.

H. Zhu, H. Tang, and T. Yang. Demand-driven Service Differentiation for Cluster-based Network Servers. In Proc. of IEEE INFOCOM'2001.

.... technologies and its costeffectiveness in achieving high availability and incremental scalability, computer clusters are increasingly recognized as the architecture of choice for scalable network services, especially when the system experiences high growth in service evolution and user demands [9, 40, 67, 85]. Within a large scale complex service cluster, service components are usually partitioned, replicated, and aggregated. Partitioning is introduced when the service processing requirement or data volume exceeds the capacity of a single server node. Service replication is commonly employed to ....

....complements them in the following three aspects. ffl Flexible resource management objectives. Most previous studies have been using a monolithic metric to measure resource utilization and define QoS constraints. Commonly used ones include system throughput, mean response time, mean stretch factor [85], or the tail distribution of the response time [60] Neptune introduces a unified metric that links the overall system efficiency with individual service response time. To be more specific, we consider the fulfillment of a service request produces certain quality aware service yield depending on ....

[Article contains additional citation context not shown here]

H. Zhu, H. Tang, and T. Yang. Demand-driven Service Differentiation for Cluster-based Network Servers. In Proc. of IEEE INFOCOM'2001.

.... and easy to manage services, the deployment of largescale complex server clusters has been rapidly emerging in which service components are usually partitioned, replicated, and aggregated [13, 14, 20, 23] Limited studies have been conducted on service differentiation for cluster based servers [27] and there is still a lack of comprehensive QoS support for these large scale cluster based network services. This paper presents the design and implementation of a flexible and efficient QoS framework for cluster based network services. This framework addresses the issues of flexible service ....

....framework should be flexible enough to allow service providers to express a variety of desired service qualities. Most previous studies have been using a monolithic metric to measure system utilization and define QoS constraints, be it system throughput, mean response time, or mean stretch factor [1, 23, 27]. On the contrary, we give service providers the flexibility of choosing the system utilization metrics that best suit their own needs or the nature of individual services. To be more specific, we consider the fulfillment of a service request generates certain yield, called QoS yield, which may be ....

[Article contains additional citation context not shown here]

H. Zhu, H. Tang, and T. Yang. Demand-driven Service Differentiation for Cluster-based Network Servers. In Proc. of IEEE INFOCOM'

....and the results of our performance evaluation. 1 Introduction High availability, incremental scalability, and manageability are some of the key challenges faced by designers of Internet scale network services and using a cluster of commodity machines is cost effective for addressing these issues [4, 7, 17, 23]. Previous work has recognized the importance of providing software infrastructures for cluster based network services. For example, the TACC and MultiSpace projects have addressed load balancing, failover support, and component reusability and extensibility for cluster based services [7, 12] ....

H. Zhu, H. Tang, and T. Yang. Demand-driven Service Differentiation for Cluster-based Network Servers. In Proc. of IEEE INFOCOM'2001, Anchorage, AK, April 2001.

No context found.

H. Zhu, H. Tang, and T. Yang, "Demand-Driven Service Differentiation in Cluster-Based Network Servers," Proc. IEEE Infocom, Apr. 2001.

No context found.

H. Zhu, H. Tang, and T. Yang. Demand-driven service differentiation for cluster-based network servers. In Proc. IEEE INFOCOM, pages 679--688, 2001.

No context found.

H. Zhu, H. Tang, and T. Yang. Demand-driven service differentiation for cluster-based network servers. In Proc. IEEE INFOCOM, pages 679--688, 2001.

No context found.

H. Zhu, H. Tang, and T. Yang. Demanddriven service differentiation in cluster-based network servers. In Proc. IEEE Infocom 2001.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC