Edith Cohen and Haim Kaplan. Prefetching the means for document transfer: A new approach for reducing Web latency. In Proc. IEEE INFOCOM, volume 2, pages 854--863, Tel-Aviv, Israel, March 2000.  @ NUS  Home/Search   Document Details and Download   Summary   Related Articles   Check

....however, prefetching during idle periods only enhances the performance since other users could not have shared the idle bandwidth anyway. Fireclick s products are based on this idea [30] Additionally, it has been found that prefetching the information about the connection can reduce web latency [31]. This approach in effect takes advantage of the locality of DNS, router and origin server memory access and reduction of TCP connection setup. 3. Content Distribution in Wireless IP Networks Contrary to the wireline web content distribution, where the emphasis is on network and server ....

E. Cohen and H. Kaplan, "Prefetching the means for document transfer: a new approach for reducing web latency," in Proceedings of INFOCOM '00, Tel Aviv, Israel, 2000.

....one transfer occurs per connection. Figure 4: Possible inter chunk delay schedules for sending to client. 3.2 DNS resolution Issue. The first request to an origin server requires a DNS lookup to determine the server s IP address. In some cases, this process can take a noticeable amount of time [8]. For subsequent HTTP requests to the same server, no lookup will be required, as the IP address will have been saved in cache. It is important that even though one proxy may have issued a request that encountered a delay because of DNS resolution times, requests by other proxies should still ....

E. Cohen and H. Kaplan. Prefetching the means for document transfer: A new approach for reducing Web latency. In Proceedings of IEEE INFOCOM, Tel Aviv, Israel, Mar. 2000.

.... as some Web requests have (potentially undesirable) side e#ects, such as adding items to an online shopping cart (which we discuss further in Chapter 12) An alternative to true prefetching has been suggested by Cohen and Kaplan when they proposed techniques that do everything but prefetch [CK00] That is, they demonstrated the usefulness of 1) pre resolving (performing DNS lookup in advance) 2) preconnecting (opening a TCP connection in advance) and 3) pre warming (sending a dummy HTTP request to the origin server) These techniques can be implemented in both proxies and browsers, and ....

.... The prefetching of content may cause problems, as not all content is cacheable (so prefetching it only wastes resources) and prefetching even cacheable content can abuse server and network resources, worsening the situation [Dav01a, CB98] An alternative is to do everything but prefetch [CK00] that is, to resolve the DNS in advance, connect 108 Unknown 11 2 3 4 7 8 15 16 31 32 63 64 127 128 255 256 511 512 1023 0.05 0.10 0.15 0.20 0.25 0.30 0.35 Distance to most recent reference to current page prefetchable Figure 6.2: The distribution of the minimum distance, ....

[Article contains additional citation context not shown here]

Edith Cohen and Haim Kaplan. Prefetching the means for document transfer: A new approach for reducing Web latency. In Proceedings of IEEE INFOCOM, Tel Aviv, Israel, March 2000.

....to variations in the wide area resolution mechanism. Some studies demonstrate that even when caching is enabled the lookup times for domain names can be quite long. Wills and Shang [10] report lookup times exceeding 2. 0 seconds for as many as 29 of lookups to random servers, and Cohen and Kaplan [14] report lookup times exceeding 3.0 seconds for as many as 10 of lookups. They also note that caching is applicable to only about 30 to 50 of requests. It is for non cached names that users will expe rience the longest lookup times. For this reason we study the behavior of DNS only for ....

....on Web latency. Their three part study involved replaying cache logs to study cache usage, measuring response time using DNS resolver routines, and measuring end to end performance of web page retrieval including DNS lookup time. This study is also performed at one location. Cohen and Kaplan [14], 24] propose and evaluate several approaches for reducing the latency of Web transfers, with particular emphasis on DNS related techniques like DNS prefetching and cache renewal policies. Cranor et al. 7] present a method for characterizing the types of remote entities sending DNS requests. ....

E. Cohen and H. Kaplan, "Prefetching the means for document transfer: A new approach for reducing web latency," in Proceedings of IEEE Infocom, 2000.

....reduces connection setup latency. Failed predictions increase the size of the cache and require new requests to resolve names that could not be predicted. A secondary prediction of aggregating requests is also possible. Two other cross domain DNS prediction examples appear in other work. In [CK00] DNS prefetching is a by product ofprefetching the means for web requests. Initiating predictive connections and sending dummy requests sets up state needed to reduce the latency for actual requests. The web browser generates predictions based on document content or user request patterns. ....

Cohen, E., and Kaplan, H., "Prefetching the Means for Document Transfer: A New Approach for Reducing Web Latency," in Proceedings of the Nineteenth Annual Joint Conference of the IEEE Computer and Communications Societies (INFOCOM '00), Tel Aviv, Isreal, March 2000, v. 2, pp. 854-863. 144

....will a#ect proxy performance [15] To add management of DNS lookups. In the current implementation the first proxy doing the lookup will pay the penalty for a DNS miss, but not subsequent proxies, thus penalizing prefetching proxies inappropriately. Since DNS resolution costs can be severe [14], this will be an important e#ect to control. We have outlined many areas above for future work. Some of these are straightforward, and their presence is just a matter of hindsight. Others, however, are more of a challenge, and may require significant e#ort. 10 Summary This technical report has ....

E. Cohen and H. Kaplan. Prefetching the means for document transfer: A new approach for reducing Web latency. In Proceedings of IEEE INFOCOM, Tel Aviv, Israel, Mar. 2000.

....advantages of such estimate: first, the server response time has no contribution at all in such RTT estimate since the ACK in response to the client initial SYN is sent by the TCP layer. Such server response time, which does not affect the RTT estimate, is rather significant for cold HTTP servers [10], i.e. servers being accessed from the client after an idle period. Even though there is also a cold route effect which does affect our RTT estimate we note from [10] that the cold server effect is significantly more important. Thus, our RTT estimate provides a RTT interval value with minimal ....

....by the TCP layer. Such server response time, which does not affect the RTT estimate, is rather significant for cold HTTP servers [10] i.e. servers being accessed from the client after an idle period. Even though there is also a cold route effect which does affect our RTT estimate we note from [10] that the cold server effect is significantly more important. Thus, our RTT estimate provides a RTT interval value with minimal error. In order to evaluate dependence of throughput versus size, connection initiation time of day and RTT estimate we choose the concept of average mutual information ....

E. Cohen and H. Kaplan. "Prefetching the means for document transfer: A new approach for reducing web latency". In IEEE INFOCOM 00, Tel Aviv, Israel, 2000.

....advantages of such estimate: first, the server response time has no contribution at all in such RTT estimate since the ACK in response to the client initial SYN is sent by the TCP layer. Such server response time, which does not affect the RTT estimate, is rather significant for cold HTTP servers [10], i.e. servers being accessed from the client after an idle period. Even though there is also a cold route effect which does affect our RTT estimate we note from [10] that the cold server effect is significantly more important. Thus, our RTT estimate provides a RTT interval value with minimal ....

....by the TCP layer. Such server response time, which does not affect the RTT estimate, is rather significant for cold HTTP servers [10] i.e. servers being accessed from the client after an idle period. Even though there is also a cold route effect which does affect our RTT estimate we note from [10] that the cold server effect is significantly more important. Thus, our RTT estimate provides a RTT interval value with minimal error. In order to evaluate dependence of throughput versus size, connection initiation time of day and RTT estimate we choose the concept of average mutual information ....

E. Cohen and H. Kaplan. "Prefetching the means for document transfer: A new approach for reducing web latency". In IEEE INFOCOM 00, Tel Aviv, Israel, 2000.

.... by Bray [7] The prefetching of content may cause problems, as not all content is cacheable (so prefetching it only wastes resources) and prefetching even cacheable content can abuse server and network resources, worsening the situation [21, 15] An alternative is to do everything but prefetch [14] that is, to resolve the DNS in advance, connect in advance to the Web server, and even warm up the Web server with a dummy request. Unknown 11 2 3 4 7 8 15 16 31 32 63 64 127 128 255 256 511 512 1023 0.20 0.30 0.35 Distance to most recent reference to current page ....

E. Cohen and H. Kaplan. Prefetching the means for document transfer: A new approach for reducing Web latency. In Proceedings of IEEE INFOCOM, Tel Aviv, Israel, Mar. 2000.

....network conditions. This places additional demands on the DNS system, since name requests must be sent more frequently, increasing load on DNS servers. This can lead to increased latency due to server loads, as well as increased probability of a dropped packet and costly DNS timeout. As shown in [5], DNS lookup can be a significant portion of web transaction latency. Both of these problems can be ameliorated by introducing multiple levels of redirection. Higher level names (e.g. m.contentdistribution.net) specify a particular network or group of networks and have a relatively long ....

E. Cohen and H. Kaplan, "Prefetching the means for document transfer: A new approach for reducing Web latency", INFOCOM 2000.

....Even with a proliferation of prefetching papers in the research community, and examples of its use in commercial products, there are a number of well known difficulties with prefetching as described in the Introduction. Some researchers even suggest pre executing the steps leading up to retrieval [12] as one way to avoid some of the difficulties of content retrieval and still achieve significant latency savings. However, as suggested above, researchers and developers are still interested in the performance enhancements possible with speculative data transfer. When prefetching (and not ....

E. Cohen and H. Kaplan. Prefetching the means for document transfer: A new approach for reducing Web latency. In Proceedings of IEEE INFOCOM, Tel Aviv, Israel, Mar. 2000.

....network conditions. This places additional demands on the DNS system, since name requests must be sent more frequently, increasing load on DNS servers. This can lead to increased latency due to server loads, as well as increased probability of a dropped packet and costly DNS timeout. As shown in [5], DNS lookup can be a signi cant portion of web transaction latency. Both of these problems can be ameliorated by introducing multiple levels of redirection. Higher level names (e.g. m.contentdistribution.net) specify a particular network or group of networks and have a relatively long ....

E. Cohen and H. Kaplan, \Prefetching the means for document transfer: A new approach for reducing Web latency", INFOCOM 2000.

.... suggests using the existing DNS zone transfer mechanism as a way to add flexible load balancing capability to a nameserver [27] Finally, some recent work has proposed new mechanisms to reduce client latency related to name resolution using prefetching or proactive cache management techniques [28], 7] This work further affirms that DNS caching plays a crucial role in determining client perceived latency. VII. CONCLUSION This paper explored two important issues related to DNSbased server selection. DNS based schemes typically disable client side caching of name resolution results, ....

Edith Cohen and Haim Kaplan, "Prefetching the means for document transfer: A new approach for reducing web latency," in Proceedings of IEEE INFOCOM, Tel Aviv, Israel, March 2000.

....from object servers (e.g. web caches) Finally, some recent work has proposed new mechanisms to reduce client latency related to name resolution. One approach is to pre resolve server names by considering the user s access history along with the links appearing on the Web page being viewed [29]. Other work proposes proactive cache management in which cached name resolution results are refreshed automatically, without waiting for a query from the client [7] This work further affirms that DNS caching plays a crucial role in determining client perceived latency. 7 Conclusion This paper ....

Edith Cohen and Haim Kaplan, "Prefetching the means for document transfer: A new approach for reducing web latency," in Proceedings of IEEE INFOCOM, Tel Aviv, Israel, March 2000. 12

....published work has been done in considering the cost of this DNS resolution in the end to end cost to retrieve Web objects from a server. The most notable piece of work is a study by Cohen and Kaplan on the potential performance improvement due to browsers or proxies pre resolving a server name [1]. This technique is studied in conjunction with other work that can be done prior to a request to reduce the overall delay of an HTTP request. Cohen and Kaplan have a related paper focusing on the impact of pro active renewal policies for DNS cache entries [2] They also investigate the effect of ....

Edith Cohen and Haim Kaplan. Prefetching the means for document transfer: A new approach for reducing web latency. In Proceedings of the IEEE Infocom 2000 Conference, Tel-Aviv, Israel, March 2000. IEEE.

....purpose. The rapid reverse caches are intended to work behind the scenes to resolve IP addresses to improve request logging at the server and help the server identify communities of interest to be used in marketing the server. Rapid reverse DNS is also used to check access permissions. Lastly, [CK] suggest DNS prefetching as a way to reduce the connection setup overhead. The authors plan to implement their system as well, but the paper describing the details is not available. They intend that pre fetch information come from in line Web links, but they do not propose aggregating the DNS ....

Cohen, E., and Kaplan, H., "Prefetching the Means for Document Transfer: A New Approach for Reducing Web Latency," in Proceedings of the IEEE Infocom 2000, Tel Aviv, Israel, March 2000, to appear.

.... [CP95] CDN 96] CY97] CMPS95] CLM99] CLLM00] Cor96] Cor98] CB95] CB96] CB97] CB98] CBC95] CJ97] Cun97] DHS93] Dan98] Dav99b] Dav99a] Dav99c] Den96] DCW96] Din96] DP96] DMP95] DFKM97a] DFKM97b] DBCK98] Duc99] DMF97] Dyk98] DJD99] CKR99] CKO99] CKZ99] CK99] CK00] EWCS96] FCAB98] FJCL99] FRC98] FRCar] Fel98] FCD 99] FVYI00] FB96] GRC97] GCR97] GCR98] GC97] Gla94] GPB98] Gri97] GB97] GPV98] GNPV98] GS96a] GS96b] GS97] HMY97] Hed98] HN96] HWMS98] HSY98] HK97] HJWC98] IKY97] IST98] JC98] JDB96] JK97] JK98] Kah97] ....

Edith Cohen and Haim Kaplan. Prefetching the means for document transfer: A new approach for reducing Web latency. In Proceedings of IEEE INFOCOM, 2000. To appear.

....GET request and that its copy was validated by the cache. Responses with code 304 are typically smaller size than responses with code 200. We can see that freshness hits with a 304 response to the client had minimal service time whereas freshness hits with 200 responses re ect RTTs Our study in [36] indicates that a rst HTTP request to a server in a time period is more likely to take longer than subsequent ones. between the cache and its clients and additional processing. The gap for freshness misses between 200 and 304 responses is also similar and re ects the additional communication ....

E. Cohen and H. Kaplan, \Prefetching the means for document transfer: A new approach for reducing web latency," in Proceedings of the IEEE INFOCOM'00 Conference, 2000.

....Test whether the cached version of the object is valid, without requesting a valid copy if it is no longer valid. Under HTTP 1.0 this is performed by issuing an If Modified Since HEAD request. HTTP 1.1 provides additional mechanisms, e.g. range requests. Policies that use 1 Our study in [35] indicates that a first HTTP request to a server in a time period is more likely to take longer than subsequent ones. pure validation requests stop renewing a copy once it is invalidated, even if its renewal credit is positive. Pure validation renewals generally use less bandwidth than ....

E. Cohen and H. Kaplan, "Prefetching the means for document transfer: A new approach for reducing web latency," in Proceedings of the IEEE INFOCOM'00 Conference, 2000.

....Studies show that when long Web waits 1 In contrast, if packet loss occurs during a TCP connection, the timeout value is set adaptively according to previous history of RTT duration, which is typically considerably shorter than the default value. do occur, DNS resolutions are a significant cause [6, 14]. Such occasional unexpected long delays significantly impact the consistency of service quality, which is often measured by extremes. High variance in connecting time was not critical for applications such as email, telnet, and FTP, that dominated the Internet when DNS specifications emerged, ....

....performance of the different polices using a heterogeneous set of proxy logs, consistently obtaining significant increase in hit rate at reasonable overhead costs. Renewal vs. Preresolving Preresolving (prefetching DNS queries) is a technique related to renewal. Preresolving was proposed in [6] as a low overhead alternative to the prefetching of documents. Preresolving applies prediction schemes (e.g. by analyzing hyperlinks or tracking access patterns) to decide which hostnames (Web servers) to preresolve. The work [6] demonstrated a potential for considerable reduction in ....

[Article contains additional citation context not shown here]

E. Cohen and H. Kaplan. Prefetching the means for document transfer: A new approach for reducing web latency. In Proceedings of the IEEE INFOCOM'00 Conference, 2000.

No context found.

Edith Cohen and Haim Kaplan. Prefetching the means for document transfer: A new approach for reducing Web latency. In Proc. IEEE INFOCOM, volume 2, pages 854--863, Tel-Aviv, Israel, March 2000.

No context found.

COHEN, E., AND KAPLAN, H. Prefetching the means for document transfer: A new approach for reducing Web latency. In Proc. IEEE INFOCOM (Tel-Aviv, Israel, Mar. 2000), vol. 2, pp. 854--863.

No context found.

E. Cohen and H. Kaplan. Prefetching the means for document transfer: A new approach for reducing web latency. In Proceedings of IEEE INFOCOM, pages 854--863, 2000.

No context found.

E. Cohen and H. Kaplan, "Prefetching the means for document transfer: a new approach for reducing web latency," Computer Networks, vol. 39, no. 4, pp. 437--455, 2002.

No context found.

E. Cohen and H. Kaplan, "Prefetching the means for document transfer: A new approach for reducing Web latency," in Proc. of IEEE INFOCOM, Tel Aviv, Israel, 2000, pp. 854--863.

No context found.

E. Cohen and H. Kaplan, "Prefetching the means for document transfer: A new approach for reducing Web latency", in Proc. of IEEE INFOCOM, Tel Aviv, Israel, 2000, pp. 854-- 863.

No context found.

Edith Cohen and Haim Kaplan, "Prefetching the means for document transfer: A new approach for reducing web latency," in INFOCOM (2), 2000, pp. 854--863.

No context found.

E. Cohen and H. Kaplan. Prefetching the means for document transfer: A new approach for reducing Web latency. Proc. IEEE INFOCOM'2000.

No context found.

E. Cohen and H. Kaplan. Prefetching the means for document transfer: A new approach for reducing web latency. In INFOCOM, volume 2, pages 854-863, Tel Aviv, Israel, Mar. 2000.

No context found.

COHEN, E., AND KAPLAN, H. Prefetching the means for document transfer: A new approach for reducing web latency. In INFOCOM (2) (2000), pp. 854--863.

No context found.

E. Cohen and H. Kaplan. Prefetching the means for document transfer: A new approach for reducing web latency. In Proceedings of IEEE INFOCOM, pages 854--863, 2000.

No context found.

E. Cohen and H. Kaplan, Prefetching the means for document transfer: A new approach for reducing Web latency. IEEE INFOCOM'00.

No context found.

E. Cohen and H. Kaplan. Prefetching the Means for Document Transfer: A New Approach for Reducing Web Latency. Proc. of the IEEE INFOCOM 2000.

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.