A. Feldmann et al. Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments. In Proc. Infocom 99. IEEE, March 1999. papers/infocom99.proxy.ps.gz  @ NUS  Home/Search   Document Details and Download   Summary   Related Articles   Check

....HttpDump [21] were one of the rst attempts to construct such specialized tools. No other tool have achieved such complete on line processing of packets. Unfortunately, their limited performance have lead the authors to give up their development, restricting them to only HTTP 1.0 transactions. BLT [2, 3] is the more promising one. It is meant to provide on line HTTP traces, related with lowerlayers events (TCP aborted connections, packet loss rate or duplicated, etc. at a high performance level. With a machine comparable to ours (a 500 MHz Alpha workstation) BLT was able to capture a 12 day ....

Anja Feldmann, Ramon Caceres, Fred Douglis, Gideon Glass, and Michael Rabinovich. Performance of Web proxy caching in heterogeneous bandwidth environments. In Proceedings of the INFOCOM '99 conference, March 1999. http://www.research.att.com/ ~anja/feldmann/papers/infocom99_proxim.ps.gz.

....when using the same average amount of open connections. Even more significant reductions, typically around 50 , are achieved on the clicks logs. The resulting performance improvement is considerable, since connection establishments induce user perceived latency and overhead [16, 31, 23] wheras large number of open connections is both detrimental to throughput [4] and is more likely to reach the server s hard set limits, causing it to refuse new connections. In Section 2 we discuss the interaction between HTTP and TCP at Web servers, and the nature of server logs data. We then ....

....termination. Busy proxy servers support TCP connections on behalf of many users; and proposals for HTTP NG [22] support connection re use across users. A recent study indicates that connection caching and re use at proxy caches may reduce user perceived latencies more than document caching [16] . Thus, proxy side connection management emerges as an important challenge. Connection management at proxy servers differs in several respects from server side management: Both request time and the time response is received are available to the HTTP session layer and can be used by the ....

A. Feldmann, R. C'aceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of Web proxy caching in heterogeneous bandwidth environments. In Proceedings of the IEEE INFOCOM'99 Conference, 1999.

....like [5 7, 2, 8] examined trac data to characterize the network, protocols or the user behavior. After the birth of the Web, lots of e ort has been devoted to study caching and content delivery architecture, analyzing traces at the application level (typically log les of web or proxy servers) [9 11]. In [12] the authors use large traces collected at the university campus at Berkeley to characterize the HTTP protocol. In [13] the authors start from data collected from a large Olympic server in 1996 to understand TCP behavior, like loss recovery eciency and ACK compression. In [14] authors ....

Feldmann, A., Caceres, R., Douglis, F., Glass, G., Rabinovich, M.: Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments. IEEE INFOCOM '99 (1999) 107-116

....requests can be carried on existing TCP connections or necessitate new connection establishment. In the latter case, additional latency is incurred (due to 3 way handshake and TCP slow start ) Studies indicate that connection establishment is a signi cant component of request response time [12] and that a short TCP connection associated with a single HTTP request response rarely gets past the slow start phase [17] With persistent HTTP [23, 14] TCP connections are kept open and re used to carry imminent future HTTP requests. Persistent connections are a default under HTTP 1.1. A ....

....or, until the connection is closed by the client or the server. With persistent connections, clients and servers keep many connections open, some for idle periods, in anticipation of future requests. Requests that utilize an already open connection bene t from reduced latency. Recent study [12] indicates that connection caching and re use at proxy caches is more e ective for reducing user perceived latency than document caching. Since TCP is implemented on top of the connection less IP, the overhead associated with keeping connection state information only exists at the two endpoints. ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of Web proxy caching in heterogeneous bandwidth environments. In Proceedings of the IEEE INFOCOM'99 Conference, 1999.

....both content and freshness misses. The content hit rate is measured per object or per byte and sometimes weighted by estimated object fetching cost. It is dictated by the available cache storage and the replacement policy used. Replacement policies for Web caches were extensively studied (e.g. [3, 4, 5, 6, 7, 8, 9, 10]) Policies that seem to perform well are Least Recently Used (lru, which evicts the least recently requested object when the cache is full) Least Frequently Used (lfu, which evicts the least frequently requested object) and Greedy Dual Size (which accounts for varying object sizes and fetching ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich, \Performance of Web proxy caching in heterogeneous bandwidth environments," in Proceedings of the IEEE INFOCOM'99 Conference, 1999.

....conferencing and is fully supported by the EMBRACE system. D. Multicast Push and Proxy Services It is well known that deploying caching proxy servers at network concentration points can decrease the average response time for retrieving documents and improve the bandwidth usage. Several studies [1][2] have shown that the number of cache hits, i.e. the number of documents that could be served straight from a proxy server s cache, amounts to 30 for sufficiently large caches. Efficiency of caching proxies can be further improved by multicasting the responses to a user query to all user ....

A. Feldmann, R Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments. In Proceedings of the IEEE INFOCOM99 Conference, March 1999.

....management, virtual group management, and multicast HTTP streaming support. I. INTRODUCTION Deploying caching proxy servers for users in local area networks or at network concentration points can decrease the average response time for retrieving documents and improve the bandwidth usage. Studies [2, 6, 7, 8] have shown that the number of cache hits, i.e. the number of documents that could be served straight from a proxy server s cache, amounts to 30 for sufficiently large caches. This means that about one in every three requested documents does not have to traverse the Internet. In many cases, ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments. In Proceedings of the IEEE INFOCOM99 Conference, March 1999.

....[21] were one of the first at tempts to construct such specialized tools. No other tool have achieved such complete on line processing of packets. Unfortunately, their limited performance have lead the authors to give up their development, restricting them to only HTTP 1.0 trans actions. BLT [2, 3] is the more promising one. It is meant to provide on line HTTP traces, related with lower layers events (TCP aborted connections, packet loss rate or duplicated, etc. at a high performance level. With a machine comparable to ours (a 500 MHz Alpha workstation) BLT was able to capture a 12 day ....

Anja Foldmann, Ramon Caceres, Fred Douglis, Gideon Glass, and Michael Rabinovich. Performance of Web proxy caching in heterogeneous bandwidth environments. In Proceedings of the INFOCOM '99 conference, March 1999. http ://www. research.att.com/ ~anj a/feldmann/pape rs/infocom99_proxim. ps. gz.

....traffic characteristics or try to offer models to represent it. These works concentrate on packets, disregarding the interaction with higher layer protocols. At the application level, after the birth of the Web, many works analyzed traces, typically log files of Web or proxy servers [9] 10] [11]. Traffic analysis at this level gives insight into applications and user behavior, but sometimes fails to correlate high level phenomena with network characteristics. Other projects analyzed the same high level phenomena using traffic traces, captured form large campus networks, like the work in ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich, "Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments," IEEE INFOCOM '99, pp. 107--116, 1999.

....is rather small, and predominantly reflects the tra#c of university users. However, with concerns over privacy, the choice was made to recruit explicit volunteers instead of surreptitiously recording the activity of all users of an existing cache or packet snooping on a network link (e.g. as in [31, 30]) This log does not distinguish users it only distinguishes clients based on IP address, and so if multiple users of this proxy operated browsers on the same machine (a possibility under UNIX) or behind the same proxy, those users requests could be interleaved. Likewise, users with dynamic ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments. In Proceedings of IEEE INFOCOM, pages 106--116, New York, Mar. 1999.

....were marked as uncacheable by clients or which responses were tagged as uncacheable or private (e.g. contained cookies) by the origin servers. Feldmann et al. have shown that these types of requests and responses are quite common and thus greatly reduce the effectiveness of proxy caching [11]. In this paper we do not consider the impact of these headers on the performance of the cache. Instead our results should be seen as motivation for using cache control headers correctly, or for developing new applications or technologies to meet the needs of users, access providers, cache ....

A. Feldmann, R. Cceres, F. Douglis, G. Glass and M. Rabinovich, "Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments", Proceedings of IEEE Infocom`99, New York, NY, pp. 107-116, March 1999.

....from a server might reflect large numbers of users, but they do not record all of the users requests, and so are of little use in evaluating client or proxy caching. In order to record transactions involving large numbers of both users and servers, researchers typically employ packet sniffers [21, 22] or proxy logs [19] However, the use of caching proxies can complicate either approach. If a sniffer is located between a caching proxy and the Internet, it will fail to record requests served from the proxy cache. The logs of a caching proxy will not suffer from this problem, but such logs do ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of Web proxy caching in heterogeneous bandwidth environments. In Proc. IEEE INFOCOM, Mar. 1999.

....User Perceived Web Performance, Web Proxy, Web Caching, Content Distribution Networks 1 Introduction user perceived Web performance. There has been extensive work evaluating the macroscopic impact of Web infrastructure for improving overall Web performance, such as caching systems (e.g. [2, 6, 7, 9, 16, 17, 19, 21, 25, 30, 32, 34]) prefetching systems (e.g. 4, 11, 15, 18, 27, 29] and to a limited extent content distribution networks (e.g. 20, 22] Surprisingly, however, there has been relatively little work characterizing the impact on Web performance as perceived by the user of various Web delivery systems. ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of web proxy caching in heterogeneous bandwidth environments. In Proc. of IEEE INFOCOM '99, March 1999.

....from a server might reflect large numbers of users, but they do not record all of the users requests, and so are of little use in evaluating client or proxy caching. In order to record transactions involving large numbers of both users and servers, researchers typically employ packet sniffers [21, 22] or proxy logs [19] However, the use of caching proxies can complicate either approach. If a sniffer is located between a caching proxy and the Internet, it will fail to record requests served from the proxy cache. The logs of a caching proxy will not suffer from this problem, but such logs do ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of Web proxy caching in heterogeneous bandwidth environments. In Proc. IEEE INFOCOM, Mar. 1999.

....Web, lots of effort has been devoted to study caching and content delivery architecture, which intrinsically are based on the deep knowledge of the traffic and user behavior. Thus many works analyze traces at the application levels, typically log files of web servers or proxy servers [9] 10] [11]. These are then very helpful understand user behavior, but less interesting from the network point of view. Many projects are instead using real traffic traces, captured form large campus networks, like the work in [12] where the authors characterize the HTTP protocol by using 2 large traces ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich, "Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments," IEEE INFOCOM '99, pp. 107--116, 1999.

....limiting the usefulness of conditional requests. Moreover, all images and 50 of the HTML responses were identical for requests with two different cookies or for a cookie and non cookie request; therefore, responses with cookies could be cached and reused for subsequent requests. Feldmann et al. [1] finds that using a proxy as a data and connection cache improves latency much more than just caching data (for both slow and fast networks) The proxy would maintain persistent connections (one or a limited number for each server) with the source web servers and reuse them for servicing requests ....

Anja Feldmann, Ramon Caceres, Fred Douglis, Gideon Glass, and Michael Rabinovich. "Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments". In Proceedings of the IEEE Infocom '99 Conference, New York, NY, March 1999. IEEE.

.... approach is to provide the cache with sufficient resources such that few replacements are needed (this is the brute force approach) Supporters of the elegant approach have utilized workload characterization to develop a number of different replacement policies [3] 9] 20] 26] Feldmann et al. [15] point out that proxies may also improve performance by caching persistent connections. 3 Methodology This section describes the methodology of the workload characterization study. Section 3.1 presents background information on the data collection site. Section 3.2 discusses the data that was ....

....modification and an aborted transfer. This means that if a file increases or decreases in size by less than 5 it is considered a modification; otherwise an abort has occurred. Using this threshold value we estimate that 10.3 of all requests in the data set were aborted. Since Feldmann et al. [15] found similar values in their study of a traditional dial up ISP environment, we believe that our approach is not unrealistic. 4 Workload Characterization In this section we present the results of our workload characterization. We analyze only the reduced data set described in Section 3.3 for ....

[Article contains additional citation context not shown here]

A. Feldmann, R. Cceres, F. Douglis, G. Glass and M. Rabinovich, "Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments", Proceedings of IEEE Infocom `99, New York, NY, pp. 107-116, March 1999.

....transfers. However, as the congestion level goes up, depending on the way bandwidth is shared, some or all flows may see poor performance, possibly leading to aborted transfers, i.e. stopped before completion, due to user impatience. Empirical evidence collected from representative servers [1], 2] 3] suggests that non negligible amounts of data may correspond to aborted transfers, e.g. 1] found that 11 of all transfers were interrupted, corresponding to 20 of the transfered volume. Users may abort their transfers, e.g. push the stop button on the a browser, for various reasons, ....

....some or all flows may see poor performance, possibly leading to aborted transfers, i.e. stopped before completion, due to user impatience. Empirical evidence collected from representative servers [1] 2] 3] suggests that non negligible amounts of data may correspond to aborted transfers, e.g. [1] found that 11 of all transfers were interrupted, corresponding to 20 of the transfered volume. Users may abort their transfers, e.g. push the stop button on the a browser, for various reasons, such as incorrect document address, long connection setup time, or poor performance during transfer. ....

A. Feldmann et.al., "Performance of web proxy caching in heterogeneous bandwidth environment," IEEE INFOCOM'99, vol. 1, pp. 107-- 116.

....and analyzing behavior for the access networks that serve as the entry point for many end users into the Internet 1 . Notable examples include studies of a building wide wireless network [1] a campus dial in modem bank [2] an ISP s cable modem population [3] and the AT T WorldNet modem bank [4]. 1 In the category of access networks we do not include campus or corporate wide, richly connected networks, such as Ethernet, and instead refer to more widely available, lower bandwidth access technologies such as dial in and cable modems. These past studies largely focus on ....

....of day, with a general shape that closely matches the Gribble data. Most of the rest of the data analysis focuses on application level characteristics, rather than session information. Feldmann et al. also collect a trace in a commercial access network, specifically the AT T WorldNet modem bank [4]. Data was collected over 12 days in mid August 1997, observing nearly 80,000 unique users with over 150,000 dial up sessions. They used the trace to drive a simulation of a web proxy, in order to assess the performance of web proxy caching. The session level characteristics of the trace are not ....

Anja Feldmann, Ramon Caceres, Fred Douglis, Gideon Glass, and Michael Rabinovich, "Performance of web proxy caching in heterogeneous bandwidth environments," in Proceedings of Infocom 99, 1999.

....the detail needed to believably estimate retrieval latency. In contrast, NCS is specifically designed to estimate the latency experienced by the user, and so includes network simulation in addition to a caching implementation. Although inspired by the caching simulators described and used in [10, 9, 7], the development of NCS has proceeded independently. While this has necessitated significant redevelopment, it has had the side benefit of being an educational process for the author. An alternative might have been to use the network simulator ns [25] and extend it appropriately for prefetching. ....

....time improvement, however, is a common justification for the use of Web caching, and is arguably the initial raison d etre for content delivery networks. As mentioned above, however, there are other simulators that estimate response times. We describe three of the most prominent ones here. PROXIM [10] is a caching and network effects simulator developed by researchers at AT T Labs. Like NCS, PROXIM accounts for TCP slow start and does not consider packet losses and their resulting effects. However, PROXIM does not including prefetching, and their paper only provides for a simple level of ....

[Article contains additional citation context not shown here]

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments. In Proceedings of IEEE INFOCOM, pages 106--116, New York, Mar. 1999.

....that the associated overheads are very small. Keywords: proxy cache, hints, Web performance 2 1. Introduction Proxy caches for large Internet Service Providers can receive high request volumes. Numerous studies show that miss ratios are often at least 40 50 , even for large proxy caches [7,9,12]. Consequently, a reduction of the cache miss overheads can result in significant throughput improvements for proxy caches experiencing high request volumes. Based on this insight and on our experience with Web server acceleration [15] this paper proposes and evaluates a new approach to ....

....than fixed mappings of objects to WP nodes[1] In contrast to the method in [17] redirection does not cause caching of multiple object replicas and is independent of client request patterns. 7. Conclusions Based on the observation that miss ratios at proxy caches are often relatively high [7,12,9], we have developed a method to improve the performance of a cluster based Web proxy by shifting some of its cache miss related functionality to be executed on a Proxy Accelerator an extended content based router implemented on an embedded system optimized for communication. Consequently, the ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, M. Rabinovich, Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments in: Proc. IEEE INFOCOM'99 (1999).

....browses, she requests multiple objects from the same server (for example, images from one or more web pages) in quick succession. Because of this, HTTP 1.1 introduced persistent connections with pipelining. This has been shown to provide significant benefits in reducing the user perceived latency [18, 8] due to temporal locality in the servers accessed by each client and reduction in the number of packet round trips between the server and the client. In CARP, each URL is redirected to a potentially different proxy. Therefore successive requests for the same remote server may go through different ....

....by the proxy for the last 30 seconds. Deciding when to close a persistent connection at the server remains a difficult problem. We have not seen definitive solutions, and different servers adopt different timeout periods. Our parameters are in ranges deemed reasonable by several researchers (e.g. [8]) Clients Proxy Bs Bs Bs Bp Bp S1 S2 S3 Figure 6: Link bandwidths in the network model for the network connecting the clients, each proxy, and the remote servers (S1; S2; S3 in this figure) 4.2 Proxy model There are two network queues at each proxy host: the server queue for requests with ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of Web proxy caching in heterogeneous bandwidth environments. In Proceedings of the INFOCOM '99, March 1999.

....web usage have become a very active area for research. Because caching and content delivery are widely considered vital to the long term viability of the web, most of the high quality data currently available is focused on providing inputs to cache or content delivery evaluations, e.g. [15, 16, 17, 19, 22, 35, 36]. For these studies, the critical data are traces or logs of URL references, typically collected at proxies or servers. There is much less data available that is focused on how web browsing behaviors by users result in the creation of network traffic. For networking studies, the critical data are ....

.... from individual segments and reconstructing HTTP protocol characteristics from TCP connections and HTTP headers (many of which we encountered with our own analysis tools) She also presents a survey of several projects in web cache design and content compression that successfully used these traces [15, 19, 20, 27]. Our approach is to use off the shelf hardware and publicly available packet capture tools. Privacy considerations limit us from capturing more than the TCP IP header, however, as we show below, with careful analysis, significant and substantial data on protocol usage and the nature, structure, ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, M. Rabinovich, Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments, Proceedings of IEEE INFOCOM `99, 1999, pp. 107-116.

....misses. The content hit rate is measured per object or per byte and sometimes weighted by estimated object fetching cost. It is dictated by the available cache storage and the replacement policy used. Replacement policies for Web caches were extensively studied (e.g. 2] 3] 4] 5] 6] 7] [8], 9] Policies that seem to perform well are Least Recently Used (LRU, which evicts the least recently requested object when the cache is full) Least Frequently Used (LFU, which evicts the least frequently requested object) and GreedyDual Size (which accounts for varying object sizes and ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich, "Performance of Web proxy caching in heterogeneous bandwidth environments, " in Proceedings of the IEEE INFOCOM'99 Conference, 1999.

....the authors of [11] study the popularity distribution of cached static documents. The rate of change of web documents and its effect on caching is studied in [13, 16] Performance effects often overlooked in web caching, such as client side bandwidth and connection aborts are investigated in [19]. The authors of [19] conclude that caching for slow clients might in fact increase, not decrease, network bandwidth consumed by the cache as well as average client response time due to extra delays in propagating connection aborts to servers upon overload. The efforts closest to ours addressed ....

....[11] study the popularity distribution of cached static documents. The rate of change of web documents and its effect on caching is studied in [13, 16] Performance effects often overlooked in web caching, such as client side bandwidth and connection aborts are investigated in [19] The authors of [19] conclude that caching for slow clients might in fact increase, not decrease, network bandwidth consumed by the cache as well as average client response time due to extra delays in propagating connection aborts to servers upon overload. The efforts closest to ours addressed the problem of biasing ....

A. Feldmann, R. Caceres, F. Doudlis, G. Glass, and M. Rabinovich. Performance of web proxy caching in heterogeneous bandwidth environments. In Infocomm, 1999.

....the trace collection methodology. Section 5 presents the basic workload characteristics, while Section 6 focuses on our cache simulation results. Section 7 presents our stream merging results. Finally, Section 8 concludes. 2 Related Work While Web client workloads have been studied extensively [3, 5, 10, 8, 29], relatively little research has been done on multimedia traffic analysis. Acharya et al. 1] analyzed video files stored on Web servers to characterize non streaming multimedia content on the Internet. Their study showed that these files had a median size of 1.1 MB and most of them contained ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of web proxy caching in heterogeneous bandwidth environments. In Proc. of IEEE INFOCOM 1999, March 1999.

....linear approximations. From a practical standpoint, another serious weakness is that we ignore the low level aspects of Web operation. Feldmann et al. report that details such as bandwidth heterogeneity and aborted transfers can negate the bandwidth savings that proxy caching would otherwise yield [16]. An important goal of our ongoing work is to determine how much detail can be added to the model of Section 2 without sacrificing scalable decentralized computation of optimal cache sizes. The major limitation of the single cache optimization method of Section 3 is that in its present form it ....

A. Feldmann, R. Ca ceres, F. Douglis, G. Glass, M. Rabinovich, Performance of Web proxy caching in heterogeneous bandwidth environments, Proceedings of IEEE INFOCOM'99, March 1999.

....of future Web servers may look like, so that we may plan accordingly. Web server workload characterization is only one of the necessary steps for understanding the changes occurring in Web traffic. Research efforts on Web client workloads (e.g. 5] Web proxy workloads (e.g. 2] 7] 8] 15] 17][18][22] 27] 33] network traffic characterizations ( e.g. 34] as well as HTTP analyses (e.g. 4] 21] 23] 30] are all required in order better understand the Web. The remainder of the paper is organized as follows. Section 2 provides background information on the 1998 World Cup, focusing on the ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass and M. Rabinovich, "Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments", Proceedings of IEEE Infocomm `99, New York, NY, pp.107-116, March 1999.

No context found.

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of Web proxy caching in heterogeneous bandwidth environments. In Proc. of IEEE INFOCOM, pages 107--116, 1999.

No context found.

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of Web proxy caching in 11 heterogeneous bandwidth environments. In Proc. of IEEE INFOCOM, pages 107--116, 1999.

....was to design it on top of a connection oriented transport protocol such as TCP. At the same time, the current Web workload exhibits a large number of short page transfers and interactions for control purposes rather than data transfers. For example, in a trace of a large number of modem users [1], 26 of all interactions were cache validations that resulted in a not modified response. Arlitt et al. observed that even for high speed cable modem users (who intuitively would be more likely to access larger objects) and even considering only responses that did carry data ( successful ....

....growing number of proposals for using HTTP for new applications, which may increase the request size. Considering a 5 month trace from his research department (6 clients) Duchamp found in 1999 the average request size, including all headers, of 189 bytes [24] in the 1997 trace used for our study [1], 98 of all requests were GET and HEAD, which tend to be small. Flag HTTP Request . Port Number 64 80 088 Request ID 0 64 Request ID HTTP Response . a) Request (b) Response Fig. 2. DHTTP message formats. sume no possibility of a collision [25] 2 . The client must also let the ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich, "Performance of web proxy caching in heterogeneous bandwidth environments, " in INFOCOM-99, 1999.

....was to design it on top of a connection oriented transport protocol such as TCP. At the same time, the current Web workload exhibits a large number of short page transfers and interactions for control purposes rather than data transfers. For example, in a trace of a large number of modem users [15], 26 of all interactions were cache validations that resulted in a not modified response. Arlitt et al. observed that even for high speed cable modem users (who intuitively would be more likely to access larger objects) and even considering only responses that did carry data ( successful ....

....growing number of proposals for using HTTP for new applications, which may increase the request size. Considering a 5 month trace from his research department (6 clients) Duchamp found in 1999 the average request size, including all headers, of 189 bytes [14] in the 1997 trace used for our study [15], 98 of all requests were GET and HEAD, which tend to be small. 2 In fact, our DHTTP prototype uses only two byte sequence numbers for re Flag HTTP Request . Port Number 64 80 0 88 Request ID 0 64 Request ID HTTP Response . a) Request (b) Response Fig. 2. DHTTP message formats. ....

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of web proxy caching in heterogeneous bandwidth environments. In INFOCOM-99, 1999.

No context found.

A. Feldmann et al. Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments. In Proc. Infocom 99. IEEE, March 1999. papers/infocom99.proxy.ps.gz

No context found.

A. Feldmann, R. Cceres, F. Douglis, G. Glass, and M. Rabinovich, "Performance of web proxy caching in heterogeneous bandwidth environments," in Proceedings of IEEE INFOCOM '99, pp. 107-116, New York, NY, Apr. 1999.

No context found.

A. Feldmann, R. Cceres, F. Douglis, G. Glass, and M. Rabinovich, "Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments", Proceedings of the 18th Conference of the IEEE Computer and Communications Society, New York, USA, 1999.

No context found.

A. Feldmann et al., "Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments," Proc. Joint Conf. IEEE Computer and Comm. Societies (Infocom), IEEE CS Press, 1999, pp. 107--116.

No context found.

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. ael Rabinovich, "Performance of Web proxy caching in heterogeneous bandwidth environments," in Proceedings of IEEE INFOCOM '99, pp. 107--116, June 1999.

No context found.

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. ael Rabinovich, "Performance of Web proxy caching in heterogeneous bandwidth environments," in Proceedings of IEEE INFOCOM '99, pp. 107--116, June 1999.

No context found.

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of web proxy caching in heterogeneous bandwidth environments. Proc. of IEEE Conference on Computer Communications (INFOCOM'99), pp. 107-116, Mar. 1999, http://www.douglis.org/fred/work/papers/ hetproxcache.pdf.

No context found.

Anja Feldmann, Ramon Caceres, Fred Douglis, Gideon Glass, and Michael Rabinovich. Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments. In INFOCOM (1), pages 107--116,

No context found.

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich, "Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments," in Proceedings of IEEE INFOCOM '99, pp. 107--116, Mar. 1999.

No context found.

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich, "Performance of Web proxy caching in heterogeneous bandwidth environments," in Proceedings of IEEE INFOCOM '99, pp. 107--116, 1999.

No context found.

A. Feldmann, R. Cceres, F. Douglis, G. Glass, and M. Rabinovich, "Performance of web proxy caching in heterogeneous bandwidth environments," in Proceedings of IEEE INFOCOM '99, pp. 107-116, New York, NY, Apr. 1999.

No context found.

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich, "Performance of web proxy caching in heterogeneous bandwidth environment, " Proc. IEEE INFOCOM, vol. 1, pp. 107--116, 1999.

No context found.

A. Feldmann, R. Caceres, F. Douglis, G. Glass, M. Rabinovich, Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments, IEEE INFOCOM, 1999.

No context found.

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich, "Performance of Web Proxy Caching in Heterogeneous Bandwidth Environments", Proceedings of IEEE INFOCOM'99, pp. 107-116, April 1999.

No context found.

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich, "Performance of Web proxy caching in heterogeneous bandwidth environments," in Proceedings of IEEE INFOCOM '99, pp. 107--116, 1999.

No context found.

Anja Feldmann, Ramon Caceres, Fred Douglis, Gideon Glass, and Michael Rabinovich. "Performance of Web proxy caching in heterogeneous bandwidth environments." In Proceedings of the INFOCOM '99 conference, March 99.

No context found.

Anja Feldmann, Ramon Caceres, Fred Douglis, Gideon Glass, and Michael Rabinovich, Performance of web proxy caching in heterogeneous bandwidth environments, IEEE INFOCOM (1999), 107-116.

No context found.

A. Feldmann, R. Caceres, F. Douglis, G. Glass, and M. Rabinovich. Performance of web proxy caching in heterogeneous bandwidth environments. In Proceedings of the INFOCOM, 1999.

First 50 documents  Next 50

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

CiteSeer.IST at NUS - Copyright Penn State and NEC. Hosted by the School of Computing, National University of Singapore.