Y. Amir, A. Peterson, and D. Shaw. Seamlessly selecting the best copy from Internet-wide replicated web servers. In Proc. 12th Symposium on Distributed Computing, Andros, Greece, Sept. 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....the rationale for such work in terms of preliminary measurements and section 5 presents our conclusions and directions for future research. 2 Related Work or Prior Work In this section, we present some of the most relevant works in the domain of locating documents on Internet. Work done by [APS98] proposes three solutions for seamlessly locating replicas ; an HTTP, DNS and IP based schemes. The HTTP [BL95] scheme uses the HTTP redirect header to redirect a client request to the best replica; this notion of best , in this case, is relative to the application and not to the requesting ....

....on Internet, but it may be interesting for selecting the best mirror . However, it is still a global solution, as the authors suggest that all URLs be mapped in this database. The solution presented in [Pul98] implements a scheme In this case, it is the same as the DNS scheme proposed by [APS98] using the available routing information exchanged by autonomous systems (Internet Routing Register (IRR) in order to determine the shortest path the number of autonomous systems traversed for all destination IP addresses. The routing information is used by each server to build a graph linking ....

Yair Amir, Alec Peterson, and David Shaw. Seamlessly selecting the best copy from internet-wide replicated web servers. In DISC'98 - 12th International Symposium on DIStributed Computing, Andros, Greece, Sep 1998.

....such, its design goals and assumptions with respect to performance, network characteristics, security, and administration are very di erent from PAST s. More loosely related is work on overlay networks [17] ad hoc network routing [7, 22] naming [3, 9, 12, 21, 26, 29] and Web content replication [4, 18, 19]. 7. CONCLUSION We presented the design and evaluation of PAST, an Internet based global peer to peer storage utility, with a focus on PAST s storage management and caching. Storage nodes and les in PAST are each assigned uniformly distributed identi ers, and replicas of a les are stored at ....

Y. Amir, A. Peterson, and D. Shaw. Seamlessly selecting the best copy from Internet-wide replicated web servers. In Proc. 12th Symposium on Distributed Computing, Andros, Greece, Sept. 1998.

....a decision may be made quickly without further network traffic, enabling fast joins. Joining a group consists of selecting the best server and redirecting the client to that server. The details of the server selection algorithm are beyond the scope of this paper as considerable previous work [3, 18] exists in this area. Furthermore, Over cast s particular choices are constrained considerably by a desire to avoid changes at the client. Without such a constraint simpler choices could have been made, such as allowing clients to participate directly in the Overcast tree building protocol. ....

Yair Amir, Alec Peterson, and David Shaw. Seamlessly selecting the best copy from Internet-wide replicated web servers. In The 12th International Symposium on Distributed Computing (DISC'98), pages 22-23, September 1998.

....exchange limits the scalability of these routing algorithms, necessitating a hierarchical routing architecture like the one used in the Internet. Several prior works consider issues in replicating Web content in the Internet, and selecting the nearest replica relative to a client HTTP query [4, 13, 14]. Pastry provides a more general infrastructure aimed at a variety of peer to peer applications. Another related area is that of naming services, which are largely orthogonal to Pastry s content location and routing. Lampson s Global Naming System (GNS) 16] is an example of a scalable naming ....

Y. Amir, A. Peterson, and D. Shaw. Seamlessly selecting the best copy from Internet-wide replicated web servers. In Proc. 12th Symposium on Distributed Computing, Andros, Greece, Sept. 1998.

....service for mobile nodes in ad hoc wireless networks. Unlike NLS, it is designed for highly dynamic network environments of relatively small scale. Several prior works consider issues in replicating Web content in the Internet, and in selecting a nearby replica relative to a client HTTP query [10], 11] 12] 13] NLS offers a general and highly scalable solution to the problem of locating a nearby copy of an object bound to a name. The Internet DNS is a scalable naming system, but it was not designed as a location service. Although it can be manipulated to allow the resolution of a ....

Yair Amir, Alec Peterson, and David Shaw, "Seamlessly selecting the best copy from Internet-wide replicated web servers," in Proceedings of the 12th International Symposium on Distributed Computing, Andros, Greece, Sept. 1998.

....service) The document (Web Service) might reside on the server that is owned by the document publisher. It might be strategically cached by one or more servers to expedite access to the Web Service. A number of studies have investigated the impact of hosting data in close proximity to clients [23, 1]. These studies are orthogonal and can be applied to both Jini and UDDI. 3 Jini UDDI Hybrid Framework (JUHF) In Section 2, we surveyed Jini and UDDI as two complementary off the shelf frameworks for discovery and retrieval of documents in a distributed document management system. While we ....

Y. Amir, A. Peterson, and D. Shaw. Seamlessly Selecting the Best Copy from Internet-Wide Replicated Web Servers. In International Symposium on Distributed Computing, 1998.

....service for mobile nodes in ad hoc wireless networks. Unlike NLS, it is designed for highly dynamic network environments of relatively small scale. Several prior works consider issues in replicating Web content in the Internet, and selecting a nearby replica relative to a client HTTP query [15], 16] 17] 18] NLS offers a general and highly scalable solution to the problem of locating a nearby copy of an object bound to a name. The Internet DNS is a scalable naming system, but was not designed as a location service. Although it can be used to allow the resolution of a domain name ....

Yair Amir, Alec Peterson, and David Shaw, "Seamlessly selecting the best copy from Internet-wide replicated web servers," in Proceedings of the 12th International Symposium on Distributed Computing, Andros, Greece, Sept. 1998.

....a decision may be made quickly without further network traf c, enabling fast joins. Joining a group consists of selecting the best server and redirecting the client to that server. The details of the server selection algorithm are beyond the scope of this paper as considerable previous work [3, 18] exists in this area. Furthermore, Overcast s particular choices are constrained considerably by a desire to avoid changes at the client. Without such a constraint simpler choices could have been made, such as allowing clients to participate directly in the Overcast tree building protocol. ....

Yair Amir, Alec Peterson, and David Shaw. Seamlessly selecting the best copy from Internet-wide replicated web servers. In The 12th International Symposium on Distributed Computing (DISC'98), pages 22-23, September 1998.

....exchange limits the scalability of these routing algorithms, necessitating a hierarchical routing architecture like the one used in the Internet. Several prior works consider issues in replicating Web content in the Internet, and selecting the nearest replica relative to a client HTTP query [4, 13, 14]. Pastry provides a more general infrastructure aimed at a variety of peer to peer applications. Another related area is that of naming services, which are largely orthogonal to Pastry s content location and routing. Lampson s Global Naming System (GNS) 16] is an example of a scalable naming ....

Y. Amir, A. Peterson, and D. Shaw. Seamlessly selecting the best copy from Internet-wide replicated web servers. In Proc. 12th Symposium on Distributed Computing, Andros, Greece, Sept. 1998.

....such, its design goals and assumptions with respect to performance, network characteristics, security, and administration are very di erent from PAST s. More loosely related is work on overlaynetworks [17] ad hoc network routing [7, 22] naming [3, 9, 12, 21, 26, 29] and Web content replication [4, 18, 19]. 7. CONCLUSION We presented the design and evaluation of PAST, an Internet based global peer to peer storage utility,withafocus on PAST s storage management and caching. Storage nodes and les in PAST are each assigned uniformly distributed identi ers, and replicas of a les are stored at the ....

Y. Amir, A. Peterson, and D. Shaw. Seamlessly selecting the best copy from Internet-wide replicated web servers. In ##### #### ######### ## ########### #########, Andros, Greece, Sept. 1998.

....from node failures. However, Pastry and Tapestry differ in the approach they take for replicating files and in the way they achieve locality. More loosely related is work on overlay networks [18] ad hoc network routing [19, 20] naming [21, 22, 23, 24, 25, 26] and Web content replication [27, 28, 29]. 7 Conclusion We presented the design and evaluation of PAST, an Internet based global peer to peer storage utility, with a focus on PAST s storage management and caching. Storage nodes and files in PAST are each assigned uniformly distributed identifiers, and replicas of a files are stored at ....

Yair Amir, Alec Peterson, and David Shaw. Seamlessly selecting the best copy from Internet-wide replicated web servers. In Proceedings of the 12th International Symposium on Distributed Computing, Andros, Greece, September 1998.

....exchange limits the scalability of these routing algorithms, necessitating a hierarchical routing architecture like the one used in the Internet. Several prior works consider issues in replicating Web content in the Internet, and selecting the nearest replica relative to a client HTTP query [12 14]. Pastry provides a more general infrastructure aimed at a variety of peer to peer applications. Another related area is that of naming services, which are largely orthogonal to Pastry s content location and routing. Lampson s Global Naming System (GNS) 15] is an example of a scalable naming ....

Yair Amir, Alec Peterson, and David Shaw. Seamlessly selecting the best copy from Internetwide replicated web servers. In Proceedings of the 12th International Symposium on Distributed Computing, Andros, Greece, September 1998.

....paper presents a scalability analysis and an experimental evaluation of a prototype implementation of our enhanced design of such a location service. Several prior works consider issues in replicating Web content in the Internet, and selecting the nearest replica relative to a client HTTP query [3, 11, 12]. NLS offers a general and highly scalable solution to the problem of locating the nearest copy of an object bound to a name. As discussed earlier, the Internet Domain Name System is a scalable naming system that was not designed as a location service. Although it can be used to allow the ....

Y. Amir, A. Peterson, and D. Shaw. Seamlessly selecting the best copy from Internet-wide replicated web servers. In Proceedings of the 12th International Symposium on Distributed Computing, Andros, Greece, Sept. 1998.

....Language Transformation (XSLT) Along with these technologies comes the added processing required by the web server. Web servers are becoming more complex entities, they are no longer just file servers, but have the added responsibility of creating dynamic web content. We shall examine two papers [3] in section 2.1, and [10] in section 2.2, that attempt to improve response time by implementing the replicated server approach. 2.1 Seamlessly Selecting the Best Copy from Internet Wide Replicated Web Servers When server replicas were first used on the Internet, they were used solely to remedy ....

....they were used solely to remedy the second factor that was previously discussed, that of reducing the load on a single server. The traditional technique was to cluster a set of servers at the same site in an attempt to improve performance by sharing the load between several identical servers [3]. In addition to the replicated servers, paper [3] attempts to also provide assistance to the first factor in bringing the web servers closer to the client s requesting content. This technique attempts to provide this added value in a seamless manner to the user. The architecture also improves ....

[Article contains additional citation context not shown here]

Amir, Y., Peterson, A., and Shaw, D., Seamlessly Selecting the Best Copy from Internet-Wide Replicated Web Servers, Proc. 12th International Symposium on Distributed Computing (DISC'98), Andros, Greece, September 1998.

....facility that permits users to 19 perform customizations on network packets in an active network. The Metacomputing Project at JHU is exploring ways to effectively utilize unused resources available on the network to complete computation on time on networks of workstations and the global Internet [15, 14]. A commercial effort attempting to create a global ubiquitous computing framework is the HP Information Utility [80] 2.6 Formal Reasoning about Distributed Systems A number of formalisms have been developed for specifying and reasoning about concurrent systems [1, 30, 112, 113] These include ....

Yair Amir, Alec Peterson, and David Shaw. Seamlessly selecting the best copy from internet-wide replicated web servers. Technical Report CNDS-98-3, Johns Hopkins University, Baltimore, Maryland, 1998.

No context found.

A. Peterson Y. Amir and D. Shaw, "Seamlessly selecting the best copy from internet-wide replicated web servers," in International Symposium on Distributed Computing (DISC98), 1998, pp. 22--23.

....failure would result in a PANIC. 5. The Director Transparent Directing of Clients to Servers The Director is actually the standard distributed domain name system (DNS) 10, 11] used on the Internet without any changes. The Walrus system uses the DNS Round Trip Times method we proposed in [3] to transparently direct the user to the correct server. A brief recap of this method follows. A more complete description of this method and how it is used in Walrus is given in [13] The DNS Round Trip Times method takes advantage of the fact that each local name server, when querying a remote ....

....name server to settle on the quickest remote name server. To aid this process, a low time to live (TTL) value can be set on the DNS information to force it to be re requested with greater frequency. There are many other details and limitations with this method, and readers are encouraged to see [3, 13] for a more complete discussion. Phase One: Problem Notification For every server returning PANIC notify a human being. Phase Two: Locality Preference For every area whose home cluster has at least one functioning server below the threshold assign it to this home cluster. Phase Three: ....

[Article contains additional citation context not shown here]

Y. Amir, A. Peterson and D. Shaw. Seamlessly Selecting the Best Copy from Internet-Wide Replicated Web Servers. The 12th International Symposium on Distributed Computing (DISC'98) (formerly WDAG), pages 22-33, Lecture Notes in Computer Science 1499, September 1998.

....the usual statistics. PANIC similar to BYPASS, but requires notification to a human being. For example, a Web server process failure would result in a PANIC. 5. The Director Transparent Directing of Clients to Servers The Walrus system uses the DNS Round Trip Times method we proposed in [APS98] to transparently direct the user to the correct server. A brief recap of this method, which uses the domain name system (DNS) RFC 1034, RFC 1035] follows. A more complete description of this method and how it is used in Walrus is given in [Shaw98] The DNS Round Trip Times method takes advantage ....

....name server to settle on the quickest remote name server. To aid this process, a low time to live (TTL) value can be set on the DNS information to force it to be re requested with greater frequency. There are many other details and limitations with this method, and readers are encouraged to see [Shaw98, APS98] for a more complete discussion. 6. The Controller Dynamic Assignment of Areas to Clusters The Controller is the heart of the Walrus system, responsible for making the dynamic assignment of areas to clusters. For fault tolerance reasons, there exist several Controllers in the network. The ....

[Article contains additional citation context not shown here]

Y. Amir, A. Peterson and D. Shaw. Seamlessly Selecting the Best Copy from Internet-Wide Replicated Web Servers. The 12th International Symposium on Distributed Computing (DISC'98) (formerly WDAG), pages 22-33, Lecture Notes in Computer Science 1499, September 1998.

....detect their status from their position in the list. The master Controller, which is the only one to pass its decisions onto the Director, is determined by being the first group member on this list. 21 7. The Director The Walrus system uses the DNS Round Trip Times method we proposed in [APS98] to transparently direct the user to the correct server. A brief recap of this method, which uses the domain name system (DNS) RFC 1034, RFC 1035] follows: 7.1 DNS Resolution . Every server on the Internet has a fully qualified domain name (FQDN) FQDN are usually read from left to right, and ....

....or the Web server if several clients sharing a local name server all request a document from the same Web server within the time period, that still counts as one connection. 27 8. Additional Considerations 8. 1 Experience Some tests of the DNS Round Trip Times routing method were done for [APS98]. This experience produced the following recommendations: The time it takes for the DNS system to converge on the best authoritative name server increases linearly with the number of name servers. At the same time, the law of diminishing returns dictates that the added benefit derived from each ....

[Article contains additional citation context not shown here]

Amir, Y, Peterson, A, Shaw, D.: Seamlessly Selecting the Best Copy from Internet-Wide Replicated Web Servers. The 12th International Symposium on Distributed Computing (DISC'98) (formerly WDAG), Andros, Greece, September 1998, to appear.

No context found.

Y. Amir, A. Peterson, and D. Shaw. Seamlessly selecting the best copy from Internet-wide replicated web servers. In Proc. 12th Symposium on Distributed Computing, Andros, Greece, Sept. 1998.

No context found.

Y. Amir, A. Peterson, and D. Shaw. Seamlessly selecting the best copy from Internet-wide replicated web servers. In Proc. 12th Symposium on Distributed Computing, Andros, Greece, Sept. 1998.

No context found.

Yair Amir, Alec Peterson, and David Shaw. Seamlessly selecting the best copy from internet-wide replicated web servers. In DISC'98 - 12th International Symposium on DIStributed Computing, Andros, Greece, Sep 1998.

No context found.

Y. Amir, A. Peterson, and D. Shaw. Seamlessly selecting the best copy from Internet-wide replicated web servers. In Proc. 12th Symposium on Distributed Computing, Andros, Greece, Sept. 1998.

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