Jim Challenger, Paul Dantzig, and Arun Iyengar. A scalable and highly available system for serving dynamic data at frequently accessed web sites. In Proceedings of Supercomputing '98, November 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....les used in our experiments are minor modi cations of standard Webstone [12] CGI les and return a le containing randomly generated characters of the speci ed size. In the Internet the fraction of dynamic requests varies from site to site. Some sites experience more than 25 dynamic requests [7, 10]. 4.1 Overload Behaviour In this section we present some experiments that compare the overload behaviour of event driven servers to Apache. In the experiments, the server does not deploy our overload protection scheme. Requests for large static les. In the rst experiment, the sclient trac ....

Challenger J., Dantzig P., Iyengar A.: A Scalable and Highly Available System for Serving Dynamic Data at Frequently Accessed Web Sites. Proc. of ACM/IEEE SC 98 (1998)

....The dynamic les used in our experiments are minor modi cations of standard Webstone [15] CGI les and return a le containing randomly generated characters of the speci ed size. The fraction of dynamic requests varies from site to site with some sites experiencing more than 25 dynamic requests [22, 21]. For the acceptance rate of both CGI scripts and large les, minimum rates can be speci ed. The reason for this is that the processing of CGI scripts or large les should not be completely prevented even under heavy load. This minimum rate is set to 10 reqs sec in our experiments. In the next ....

Challenger J., Dantzig P., Iyengar A.: A Scalable and Highly Available System for Serving Dynamic Data at Frequently Accessed Web Sites. Proc. of ACM/IEEE SC 98 (1998)

....dynamic content is distributed across three different servers: the Web server, the application server, and the database management server. Consequently, it is not straightforward to create a mapping between the data and the corresponding Web pages automatically in contrast to the other approaches [3, 4, 5, 6], which assume that such map pings are provided by the system designers. The detailed descriptions on automated construction of such URL and database query mapping are given in [7] and Section 3. 2.3 Redundant Web Application Server Ap proach Traffic Director] Figure 3: System Architecture ....

....from Persistence Software is one of the first dynamic caching solution that is available as a product. However, Dynamai relies on proprietary software for both database and application server components. Thus it cannot be easily incorporated in exist ing e commerce framework. Challenger et al. [4, 5, 6] at IBM Research have developed a scalable and highly available system for serving dynamic data over the Web. In fact, the IBM system was used at Olympics 2000 to post sport event results on the Web in timely Figure 14: Performance matrix measurement for various caching solution deployments (cache ....

Jim Challenger, Paul Dantzig, and Arun Iyengar. A Scalable and Highly Available System for Serving Dynamic Data at Frequently Accessed Web Sites. In Proceedings of ACM/IEEE Supercom- puting'98, Orlando, Florida, November 1998.

....2 gives the optim value of fyo, 4 Evaluation 4.1 Methodology We constructed workloads using web server logs obtained from several production Internet servers. The first workload, Olympics98, is derived from the requests received on Feb 19, 1998 at the Nagano Winter Olympics servers at Columbus [2]. The second workload, Finance, is derived from the requests recorded on Oct 19, 1999 at the web site of a major financial services company. Avg requests sec 97 16 Peak requests sec 171 46 Avg requests corm 12 8.5 Files 61,807 16,872 Total file size 705 MB 171 MB Requests 8,370,093 ....

Jim Challenger, Paul Dantzig, and Arun Iyengar. A scalable and highly available system for serving dynamic data at frequently accessed web sites. In Proceedings of 1998 ACM/IEEE Supercomputing (SC98), Orlando, Florida, November 1998, 1998.

....a proxy from the caching neighbourhood. By inducting caches into the neighbourhood, the server authorizes these caches to deliver content on its behalf. One difference between domain caching and this approach is that in domain caching, all clients can volunteer to act as caches. Challenger et al. [5] describe a technique for serving dynamic data at busy web sites. This approach can be used by web sites that dynamically generate content in response to changes in data values (For example, generating a web page when the stock value of a company changes) They construct a graph that captures ....

J. Challenger, P. Dantzig, and A. Iyengar. A scalable and highly available system for serving dynamic data at frequently access web sites. In In Proceedings of the SC98, Nov 1998.

....a proxy from the caching neighbourhood. By inducting caches into the neighbourhood, the server authorizes these caches to deliver content on its behalf. One difference between domain caching and this approach is that in domain caching, all clients can volunteer to act as caches. Challenger et al. [4] describe a technique for serving dynamic data at busy web sites. This approach can be used by web sites that dynamically generate content in response to changes in data values (For example, generating a web page when the stock value of a company changes) They construct a graph that captures ....

Jim Challenger, Paul Dantzig, and Arun Iyengar. A scalable and highly available system for serving dynamic data at frequently access web sites. In In Proceedings of the SC98, Nov 1998.

....cache priority scheme to account for the di erent levels of request priority that a web site might like to support. 3 Related Work 3.1 Keeping caches consistent The following papers address the issue of maintaining caches for dynamic data consistent. The Data Update Propagation (DUP) algorithm [3, 4] maintains an object dependency graph (ODG) that links the underlying data with the objects (web pages) dependent on the data. Challenger et al. discuss the application of DUP to enabling a cache invalidation scheme, where the invalidations can be generated as a result of database triggers. The ....

Jim Challenger, Paul Dantzig, and Arun Iyengar. A Scalable and Highly Available System for Serving Dynamic Data at Frequently Accessed Web Sites. In In Proceedings of Supercomputing '98, 1998.

No context found.

J. Challenger, P. Dantzig, and A. Iyengar. A Scalable and Highly Available System for Serving Dynamic Data at Frequently Accessed Web Sites. In Proceedings of ACM/IEEE, Supercomputing '98 (SC98), November 1998.

No context found.

J. Challenger, P. Dantzig, and A. Iyengar. A Scalable and Highly Available System for Serving Dynamic Data at Frequently Accessed Web Sites. In Proceedings of ACM/IEEE SC98, November 1998.

No context found.

J. Challenger, P. Dantzig, and A. Iyengar. A Scalable and Highly Available System for Serving Dynamic Data at Frequently Accessed Web Sites. In Proceedings of ACM/IEEE SC98, November 1998.

No context found.

Paul Dantzig Jim Challenger and Arun Iyengar. A scalable and highly available system for serving dynamic data at frequently accessed web sites. In In Proceedings of ACM/IEEE, Supercomputing '98 (SC98), 1998.

No context found.

J. Challenger, P. Dantzig, and A. Iyengar. A Scalable and Highly Available System for Serving Dynamic Data at Frequently Accessed Web Sites. In In Proceedings of ACM/IEEE, Supercomputing '98 (SC98), November 1998.

....cached data. Consequently, we allow dynamic Web pages to be cached as well as static ones, since applications can explicitly invalidate any page whenever it becomes obsolete. Caching of dynamic Web pages is essential for improving the performance of Web sites containing significant dynamic content [2,3,14,15]. Multiprocessor accelerators can further increase the performance. Our multiprocessor system ar chitecture consists of a cluster of Web accelerator cache nodes and a front end load balancer. From a scalability standpoint, the objective is to combine the individual cache space of each member of ....

....data is typically serviced by returning a file. Web servers often consume several orders of magnitude more CPU time creating a dynamic page than a comparably sized static page. For Web sites containing significant dynamic content, it is essential to cache dynamic pages to improve performance [2,3,14,15]. We are not aware of any Web server accelerator besides our own which allows dynamic pages to be cached. All cached data must be stored in memory. Caching objects on disk would slow down the accelerator too much. Consequently, cache sizes are limited by memory sizes. Our accelerator uses the ....

J. Challenger, P. Dantzig, A. Iyengar, A scalable and highly available system for serving dynamic data at frequently accessed web sites, Proceedings of ACM/IEEE SC98, November 1998.

.... improve the scalability of lease systems by forming a distribution tree for invalidations and by serving renewal requests from lower level caches [25] Third, reverse proxy caching of dynamically generated data at the server can achieve nearly 100 hit rates and can dramatically reduce server load [3]. By implementing our system as a hierarchy, we make it easy to gain these advantages. Further, if a multi level hierarchy is used (such as the Squid regional proxies [18] or a cache mesh [20] we speculate that nodes higher in the hierarchywillachieve hit rates between the per client cache hit ....

J. Challenger, P. Dantzig, and A. Iyengar. A Scalable and Highly Available System for Serving Dynamic Data at Frequently Accessed Web Sites. In ACM/IEEE SC98,November 1998.

.... improve the scalability of lease systems by forming a distribution tree for invalidations and by serving renewal requests from lower level caches [25] Third, reverse proxy caching of dynamically generated data at the server can achieve nearly 100 hit rates and can dramatically reduce server load [3]. By implementing our system as a hierarchy, we make it easy to gain these advantages. Further, if a multi level hierarchy is used (such as the Squid regional proxies [18] or a cache mesh [20] we speculate that nodes higher in the hierarchy will achieve hit rates between the per client cache hit ....

J. Challenger, P. Dantzig, and A. Iyengar. A Scalable and Highly Available System for Serving Dynamic Data at Frequently Accessed Web Sites. In Proceedings of ACM/IEEE SC98, November 1998.

....to serve the pages statically, it may not be feasible to keep the files current. This is particularly true if the number of files which need to be updated frequently is large. Consequently, the official Web site for the 1998 Olympic Winter Games generated a high percentage of Web pages dynamically [3]. Whenever new content became available to the computers implementing the Web site, updated Web pages reflecting these changes were made available to the rest of the world within seconds. Clients could thus rely on the Web site to provide the latest results, news, photographs, and other ....

....site are significantly less than those required for proxy caches which store data from several Web sites [5, 2, 8] The maximum memory required for a single copy of all cached objects was around 175 Mbytes. A detailed description of the architecture of the Olympic Games Web site is contained in [3]. A total of 634.7 million requests were serviced during the Olympic Games. On the peak day (Day 7, Feb 13) the site served 56.8 million requests over a 24 hour period. By contrast, the 1996 Olympic Games Web site peaked at 17 million hits a day, fewer than any day for the 1998 Olympic Games. ....

[Article contains additional citation context not shown here]

J. Challenger, P. Dantzig, and A. Iyengar. A Scalable and Highly Available System for Serving Dynamic Data at Frequently Accessed Web Sites. In Proceedings of ACM/IEEE SC98, November 1998.

No context found.

Jim Challenger, Paul Dantzig, and Arun Iyengar. A scalable and highly available system for serving dynamic data at frequently accessed web sites. In Proceedings of Supercomputing '98, November 1998.

No context found.

Jim Challenger, Paul Dantzig, and Arun Iyengar. A scalable and highly available system for serving dynamic data at frequently accessed web sites. In Proceedings of the 1998.

No context found.

J. Challenger, P. Dantzig, and A. Iyengar. A scalable and highly available system for serving dynamic data at frequently accessed web sites. In Proceedings of Supercomputing'98, Nov. 1998.

No context found.

Jim Challenger, Paul Dantzig, and Arun Iyengar. A Scalable and Highly Available System for Serving Dynamic Data at Frequently Accessed Web Sites. In Proceedings of Supercomputing'98, 1998.

No context found.

Jim Challenger, Paul Dantzig, and Arun Iyengar. A scalable and highly available system for serving dynamic data at frequently accessed web sites. In Proceedings of Supercomputing '98, November 1998.

No context found.

J. Challenger, P. Dantzig, and A. Iyengar. A scalable and highly available system for serving dynamic data at frequently accessed web sites. In Proceedings of Supercomputing'98, Nov. 1998.

No context found.

J. Challenger, P. Dantzig, and A. Iyengar. A scalable and highly available system for serving dynamic data at frequently accessed web sites. In In Proceedings of ACM/IEEE, Supercomputing '98 (SC98), Nov. 1998.

No context found.

Jim Challenger, Paul Dantzig, and Arun Iyengar. A Scalable and Highly Available System for Serving Dynamic Data at Frequently Accessed Web Sites. In Proceedings of Supercomputing'98, 1998.

No context found.

J. Challenger, P. Dantzig, et al. A scalable and highly available system for serving dynamic data at frequently accessed web sites. In ACM/IEEE, Supercomputing '98. Nov. 1998.

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