D. Andresen, T. Yang, V. Holmedahl, and O. Ibarra. SWEB: Towards a Scalable WWW Server on MultiComputers. In Proc. of the IEEE Intl. Symposium on Parallel Processing, pages 850--856, Apr. 1996.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....phase. In all cases, the provider s revenue is maximized. 6. Related Work Our work is closely related to four groups of previous work: large scale resource management infrastructures [22, 11, 23] SLA enforcement in server farms [1, 13] request schedulers for cluster based network services [9, 10, 12, 14, 20, 28, 25, 29], and queue based network QoS algorithms [19, 27, 24, 21] Grid infrastructures such as Globus [22] Legion [11] and Condor [23] need to respect sharing agreements when pooling together resources belonging to multiple domains. However, existing solutions in these infrastructures have ei ....

....Oc eano like systems are specified. Cluster based network servers are now widely used. Commercial products [9, 10] are available to be used as front ends for cluster servers. Most request distribution strategies fall into two categories: weighted round robin and its variations for load balancing [12, 14, 20, 28], and content aware load distribution for locality [25, 29] Our work focuses on an orthogonal problem, of ensuring that requests from different customers receive service commensurate with pre specified agreements. More related to our work is the Cluster Reserves technique [13] a mechanism to ....

D. Andresen, T. Yang, V. Holmedahl, and O. Ibarra. SWEB: Towards a scalable WWW server on multicomputers. In Proceedings of the 10th IPPS, 1996.

....load across multi server homogeneous web sites by leveraging the DNS service used to provide the mapping between a web page s URL and the IP address of a web server serving the URL. Round robin DNSwas proposed, where the DNS system maps requests to web servers in a round robin fashion [KBM94] AYHI96] Because DNS mappings have a Timeto Live (TTL) field associated with them and tend to be cached at the local name server in each domain, this approach can lead to a large number of client requests from a particular domain getting mapped to the same web server during the TTL period. Thus, ....

D. Andresen, T. Yang, V. Holmedahl, and O.H. Ibarra. SWEB: Towards a Scalable WWW Server on MultiComputers. In IEEE International Symposium on Parallel Processing, April 1996.

....a cluster of workstations. the nodes when another node will be responsible for serving the request. The traditional network servers distribute requests based solely on their load metric, which is usually the resource utilization or the number of open connections being handled by each node (e.g. [12, 22, 2, 14, 7, 26]) As a result, these traditional servers may suffer from high cache miss rates (when the server s working set size exceeds the size of each local memory) among other problems. To reduce miss rates, the network server can distribute requests based on the content requested and on cache locality ....

D. Andresen, T. Yang, V. Holmedahl, and O. H. Ibarra. SWEB: Towards a Scalable WWW Server on Multicomputers. In Proceedingsof the 10th IPPS, April 1996.

....placed on popular Web sites and the growing importance of electronic commerce has precipitated the need for high performance, cost e ective, and highly available server systems. Clusters of commodity workstations are becoming increasingly popular for setting up server systems in such environments [50, 7, 38]. The principal reasons for this are: A cluster of cheap workstations tends to be more cost e ective than a single workstation that can a ord performance comparable to the collective performance of the cluster. A cluster a ords incremental scalability; that is, the capacity of a cluster can ....

....Resource Management Framework for Predictable Quality of Service As the World Wide Web experiences increasing commercial and mission critical use, users and content providers expect high availability and predictable performance. Towards this end, work is being done in scalable Web infrastructure [112, 50, 7], Web content caching [87, 30, 108, 28] provision of di erentiated services in the Internet, and di erentiated services in Web servers and proxies [12, 17, 72, 26, 22, 69] This chapter focuses on the provision of predictable quality of service in Web servers and proxies. Web site and proxy ....

D. Andresen et al. SWEB: Towards a Scalable WWW Server on MultiComputers. In Proccedings of the 10th International Parallel Processing Symposium, Apr. 1996.

....The scalable servers described in [4] also apply a weighted round robin distribution but do so in a distributed and thus more fault tolerant fashion. None of these proposals takes the requests contents into account. Exceptions are the Resonate Central Dispatch product [21] and the SWEB project [1] that implement a content based request distribution. However, even these latter systems do not consider cache locality when distributing requests. In contrast, the servers of Holmedahl et al. and Pai et al. do consider cache locality. Holmedahl et al. use cooperative (disk) caching of ....

D. Andresen,T. Yang, V. Holmedahl, and O. H. Ibarra. SWEB: Towards a Scalable WWW Server on Multicomputers. In Proceedingsof the 10th International Parallel Processing Symposium, April 1996.

....connection accepted by one of the nodes when another node will be responsible for serving the request. The traditional network servers distribute requests based solely on their load metric, which is usually the resource utilization or the number of open connections being handled by each node (e.g. [9, 14, 1, 11, 5, 18]) As a result, these traditional servers may suffer from high cache miss rates (when the server s working set size exceeds the size of each local memory) among other problems. To reduce miss rates, the network server can distribute requests based on the content requested and on cache locality ....

D. Andresen, T. Yang, V. Holmedahl, and O. H. Ibarra. SWEB: Towards a Scalable WWW Server on Multicomputers. In Proceedings of the 10th IPPS, April 1996.

....servers described in [7] also apply a weighted round robin distribution but do so in a distributed and thus more fault tolerant fashion. None of these systems takes the requests contents into account. In contrast to these systems, the Resonate Central Dispatch product [26] the SWEB project [2], the master slave system proposed in [29] and the server studied in [27] all implement content based request distribution. The Central Dispatch, for instance, uses contentbased distribution to segregate requests for different types of content, such as images, text, or CGI scripts. However, even ....

D. Andresen, T. Yang, V. Holmedahl, and O. H. Ibarra. SWEB: Towards a Scalable WWW Server on Multicomputers. In Proceedingsof the 10th International Parallel Processing Symposium, April 1996.

....scalable network servers that perform content based request distribution. Web servers based on clusters of workstations PCs are widely used [18] Most commercial Web switch products for cluster servers use a request distribution strategy that does not require examining the content of the request [2, 20, 14, 9]. The most common such technique used for request distribution is some variant of weighted round robin. Resonate, Inc. 27] is an exception in that their product o ers content aware request distribution using a method similar to TCP hando . Fox et al. 18] describe a layered architecture for ....

D. Andresen et al. SWEB: Towards a Scalable WWW Server on MultiComputers. In Proccedings of the 10th International Parallel Processing Symposium, Honolulu, HI, Apr. 1996.

....Dispatch product by Resonate, Inc. supports a strategy that is also based on content based request distribution [21] Looselycoupled distributed servers are also widely deployed in the Internet and use various techniques for load balancing including DNS round robin [8] HTTP client re direction [1], Smart clients [27] source based forwarding [12] and hardware translation of network addresses [11] Load balancing strategies that seek to satisfy response time based performance criteria are discussed in [22] and the citations therein. Our work focuses on resource management in cluster based ....

D. Andresen et al. SWEB: Towards a Scalable WWW Server on MultiComputers. In Proccedings of the 10th International Parallel Processing Symposium, Apr. 1996.

....The scalable servers described in [4] also apply a weighted round robin distribution but do so in a distributed and thus more fault tolerant fashion. None of these proposals takes the requests contents into account. Exceptions are the Resonate Central Dispatch product [21] and the SWEB project [1] that implement a content based request distribution. However, even these latter systems do not consider cache locality when distributing requests. In contrast, the servers of Holmedahl et al. and Pai et al. do consider cache locality. Holmedahl et al. use cooperative (disk) caching of ....

D. Andresen, T. Yang, V. Holmedahl, and O. H. Ibarra. SWEB: Towards a Scalable WWW Server on Multicomputers. In Proceedings of the 10th International Parallel Processing Symposium, April 1996.

No context found.

D. Andresen, T. Yang, V. Holmedahl, and O. Ibarra. SWEB: Towards a Scalable WWW Server on MultiComputers. In Proc. of the IEEE Intl. Symposium on Parallel Processing, pages 850--856, Apr. 1996.

No context found.

D. Andresen, T. Yang, V. Holmedahl, and O. Ibarra. SWEB: Towards a Scalable WWW Server on MultiComputers. In Proc. of the 10th IEEE Intl. Parallel Processing Symposium, Honolulu, HI, Apr. 1996.

.... 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 ....

....architecture serves as the basis for further clustering supports including data replication, service discovery, load balancing, resource management, failure detection and recovery. Previous research has proposed and evaluated various load balancing policies for cluster based distributed systems [9, 15, 28, 32, 43, 57, 61, 82, 83]. These studies are mainly focused on coarse grain computation and they often ignore fine grain jobs by simply processing them locally. In the context of network services, with the trend toward delivering more feature rich services in real time, large number of fine grain sub services need to be ....

[Article contains additional citation context not shown here]

D. Andresen, T. Yang, V. Holmedahl, and O. Ibarra. SWEB: Towards a Scalable WWW Server on MultiComputers. In Proc. of the 10th IEEE Intl. Parallel Processing Symposium, pages 850--856, Honolulu, HI, April 1996.

....Ethernet (layer 2) or IP (layer 3) switches, which do not provide any TCP level traffic statistics. This constraint calls for more complex load information dissemination schemes. Previous research has proposed and evaluated various load balancing policies for cluster based distributed systems [6, 8, 11, 12, 17, 23, 24, 29, 30]. Load balancing techniques in these studies are valuable in general, but not all of them can be applied for cluster based network services. This is because they focus on coarse grain computation and often ignore fine grain jobs by simply processing them locally. For example, the job trace used in ....

D. Andresen, T. Yang, V. Holmedahl, and O. Ibarra. SWEB: Towards a Scalable WWW Server on MultiComputers. In Proc. of the IEEE Intl. Symposium on Parallel Processing, pages 850--856, Apr. 1996.

....throughput, and (3) handling server unresponsiveness gracefully. We also demonstrate the (4) ease ofuse of the DAC primitive and (5) the scalability of our architecture design. 1. INTRODUCTION Computer clusters are widely deployed to deliver highly scalable and available online services [2, 7, 15, 17]. Well known Web sites such as Yahoo or MSN employ service clusters with thousands of machines. The persistent data for cluster based Internet services is often partitioned and the aggregation of data produced from multiple partitions is a commonly requested operation. For instance, an Permission ....

D. Andresen, T. Yang, V. Holmedahl, and O. Ibarra. SWEB: Towards a Scalable WWW Server on MultiComputers. In IEEE IPPS, Honolulu, HI, Apr. 1996.

....5 evaluates the performance impact of real systems that simulations cannot capture. Section 6 discusses related work and Section 7 concludes the paper. 2 Evaluation Methodology Previous research has proposed and evaluated various load balancing policies for cluster based distributed systems [1, 6, 7, 11, 16, 17, 26, 27]. Load balancing techniques in these studies are valuable in general, but not all of them can be applied for cluster based network services. This is because they focus on coarse grain computation and often ignore fine grain jobs by simply processing them locally. For example, the job trace used in ....

D. Andresen, T. Yang, V. Holmedahl, and O. Ibarra. SWEB: Towards a Scalable WWW Server on MultiComputers. In Proc. of the IEEE Intl. Symposium on Parallel Processing, pages 850--856, April 1996.

....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] ....

D. Andresen, T. Yang, V. Holmedahl, and O. Ibarra. SWEB: Towards a Scalable WWW Server on MultiComputers. In Proc. of the IEEE Intl. Symposium on Parallel Processing, pages 850--856, April 1996.

....differentiation while maximizing resource utilization and that it can substantially outperform static server partitioning. I. INTRODUCTION The deployment of cluster based network services keeps increasing to meet the demand for scalability and availability for the fast growing client population [1], 2] 3] 4] However, it is not always possible for a Web site to accurately predict peak load and prepare enough computing resources because client request rates tend to be bursty and fluctuate dramatically [5] 6] Such fluctuations might not only be caused by people s working habit, or ....

....sites. Further, we focus on service differentiation in the context of server clusters. Server clustering is an effective approach for Web sites with CPU or I O bottleneck. Some previous work has studied issues related to server clustering, such as architecture design and performance optimization [1], 2] 3] 12] 13] But not many results are available on service differentiation in cluster based network servers. Service differentiation for static Web pages is studied in [14] for single server request scheduling. They do not deal with cluster based services and do not address performance ....

Dan Andresen, Tao Yang, Vegard Holmedahl, and Oscar H. Ibarra, "SWEB: Towards a Scalable WWW Server on MultiComputers," in Proceedings of the IEEE International Symposium on Parallel Processing, Apr. 1996, pp. 850--856.

....differentiation while maximizing resource utilization and that it can substantially outperform static server partitioning. I. INTRODUCTION The deployment of cluster based network services keeps increasing to meet the demand for scalability and availability for the fast growing client population [1], 2] 3] However, it is not always possible for a Web site to accurately predict peak load and prepare enough computing resources because client request rates tend to be bursty and fluctuate dramatically [4] 5] Such fluctuations might not only be caused by people s working habit, or users ....

....of CPU and disk I O capacities instead of network bandwidth. Many e commerce or e service sites rely heavily on database operations and dynamic page generations. CPU and I O resources are usually the major bottlenecks for those sites instead of network bandwidth. Previous work on server clustering [1], 2] 3] 12] 13] has addressed architecture design and performance optimization, but not service differentiation. The industry has realized the potential server resource bottleneck problem and several companies are already providing content aware switches which can be used for service ....

Dan Andresen, Tao Yang, Vegard Holmedahl, and Oscar H. Ibarra, "SWEB: Towards a Scalable WWW Server on MultiComputers, " in Proceedings of the IEEE International Symposium on Parallel Processing, Apr. 1996, pp. 850--856.

....therefore multiple clustered ADL servers can be viewed as a web server group. Previous work has addressed issues related to building web server clusters. The first workstation based Web server cluster [18] used a DNS round robin rotation to assign requests to the server nodes. The SWEB project [4] further optimized such a system by redirecting requests based on dynamic system loads. Web request redirection based on server CPU loads has also been recently studied in [8, 24] Previous load balancing research (e.g [25] normally used only one load index such as CPU usage or ready queue length ....

....project developed a prediction based scheduling scheme for assigning Web requests. The scheme attempts to predict the absolute processing time for each request based on its computation and I O profile as well as the CPU and I O load on each node. For file retrieval and image processing requests [4, 5] which have easy to identify CPU and I O usage patterns, this scheme has been shown to be very effective. It is difficult, however, to predict the cost of database queries which dominate ADL request activities. Although some research work on predicting the result size of a query is available (e.g. ....

[Article contains additional citation context not shown here]

D. Andresen, T. Yang, V. Holmedahl, and O. Ibarra. SWEB: Towards a Scalable WWW Server on MultiComputers. Proc. of Intl. Symp. on Parallel Processing, IEEE, pages 850--856, Apr. 1996.

....popular at many Web sites since it can provide new types of services such as electronic commerce and personalized information presentation. There are a number of techniques proposed to speedup dynamic page delivery, for example, efficient OS support [4, 22] server clustering and load balancing [3, 9, 12, 29], content preprocessing [10] and caching dynamic content at server and proxy sites [14, 15, 24, 25] This paper focuses on caching and invalidating dynamic pages at the server sites. Previous work on Web caching and invalidation such as [7, 13] at proxy servers mainly deals with static pages. ....

....precomputing only when computation power is available. Similar techniques to precomputing are static file prefetching which may happen between proxies and servers [17] or between clients and proxies [11] Precomputing can be viewed as a technique orthogonal to server clustering with load balancing [3, 9, 21, 30] in that server clustering distribute workloads spatially while precomputing tries to distribute load temporally. Recent invalidation techniques for Web caching use multicast networks [18, 27, 28] for efficiently enforcing consistency over the Internet. Their work does not explicitly address ....

D. Andresen, T. Yang, V. Holmedahl, and O. Ibarra. SWEB: Towards a Scalable WWW Server on MultiComputers. In Proceedings of the IEEE International Symposium on Parallel Processing, pages 850--856, Apr. 1996.

....become popular at many Web sites since it enables new types of services such as electronic commerce and personalized information presentation. A number of techniques have been proposed to speed up dynamic page delivery, including efficient OS support [5, 24] server clustering and load balancing [2, 11, 15, 31], content preprocessing [12] and dynamic content caching at server and proxy sites [17, 18, 26, 27] This paper focuses on caching and invalidating dynamic pages at the server sites. Previous work on Web caching and invalidation [8, 16] at proxy servers mainly deals with static pages. Caching ....

....URL classes to cached pages. Another difference between our work and the work in [10] is in precomputing. the work in [10] uses instant precomputing while we use selective precomputing. Selective precomputing can be viewed as a technique orthogonal to server clustering with load balancing [2, 11, 23, 32] because server clustering distributes and balances load spatially while selective precomputing tries to distribute and balance load temporally. A disadvantage of server site dynamic content caching is that every request has to go to the origin server, even though a request can use ....

D. Andresen, T. Yang, V. Holmedahl, and O. H. Ibarra. SWEB: Towards a Scalable WWW Server on MultiComputers. In Proceedings of the IEEE International Symposium on Parallel Processing, pages 850--856, Apr. 1996.

....first proposed a clustering technique that uses DNS rotation to balance load among cluster nodes. Research has demonstrated that DNS round robin rotation does not evenly distribute the load among servers, due to non uniform resource demands of requests and DNS entry caching. A number of projects [5, 9, 19] have proposed methods for more fairly distributing load among a group of servers based on HTTP redirection or intelligent DNS rotation. The main weakness of a DNS based server cluster is that the IP addresses of all nodes in a cluster are exposed to the Internet. When one server goes down for ....

....and reservation based scheduling which considers both I O and CPU utilization. We have shown that these schemes can effectively achieve load re balancing and the remote execution overhead is not only negligible but even smaller than standard local CGI execution. We do not use HTTP redirection [5] for request re scheduling because it adds client round trip latency for every rescheduled request and also exposes IP addresses of server nodes. The work on a layered architecture for network services in [15] does not have a detailed study on performance optimization and evaluation for issues ....

[Article contains additional citation context not shown here]

D. Andresen, T. Yang, V. Holmedahl, and O. Ibarra. Sweb: Towards a scalable WWW server on multicomputers. Proc. of Intl. Symp. on Parallel Processing, IEEE, pages 850--856, April 1996.

No context found.

D. Andresen et al. SWEB: Towards a Scalable WWW Server on MultiComputers. In Proccedings of the 10th International Parallel Processing Symposium, Apr. 1996.

No context found.

D. Andresen et al. SWEB: Towards a Scalable WWW Server on MultiComputers. In Proccedings of the 10th International Parallel Processing Symposium, Apr. 1996.

First 50 documents

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