Weisong Shi, Eli Collins, and Vijay Karamcheti. Modeling Object Characteristics of Dynamic Web Content. Special Issue on scalable Internet services and architecture of Journal of Parallel and Distributed Computing (JPDC), Sept. 2003.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....based on known relationships between objects instead of using heuristics to infer relationships. Shi et al. also download pages from a set of Web sites at regular intervals. They use the collected data to analyze page structure and derive models that characterize dynamically generated content [9]. 7. SUMMARY This work examines a new policy, MONARCH, for strong cache consistency on the Web and uses a novel evaluation method to compare the performance of this policy with current and proposed policies. One contribution of the work is the evaluation of MONARCH, which combines object ....

....more realistic composition of objects and change characteristics, including more realistic sharing of objects across pages, than synthetic content. This better understanding of page composition can lead to better synthetic content generators, and steps are already being taken in that direction [9]. The simulation also allows us to compare the relative overhead of strongly consistent cache policies, although each policy evaluated incurs different types of overhead. The Always Validate policy can generate a large number of requests, the Object and Volume Lease policy requires the server to ....

W. Shi, E. Collins, and V. Karamcheti. Modeling Object Characteristics of Dynamic Web Content. In Proceedings of the IEEE Globecom, Taipei, Taiwan, Nov. 2002.

.... for reuse at the sub document level (at the granularity of individual objects making up the overall document) Two recent studies have shown that approximately 60 of the bytes in dynamic responses from a set of popular web sites could in fact be reused from a previous retrieval of the page [34, 40]. Although encouraging, the above proposals and studies need to be supplemented with characterizations of the actual workload encountered on sites that serve dynamic and personalized content. These characterizations serve two roles: first, they provide evidence for whether or not object ....

....of the distribution of channel size, where 99 of channels are less than 3000 bytes. We found the channel sizes are best modeled using a Weibull distribution (with CDF F (X) 1 e (0.0012x) 1. 6 ) This observation is in agreement with a previous study on six news and e commerce web sites [34]. It is interesting to compare this distribution with the overall document size distribution in Figure 7(b) The latter shows that 70 of the documents lie in a very small range between 9,725 and 10,688 bytes. The popularity of the HOME tab and the fact that the layout accounts for a sizeable ....

[Article contains additional citation context not shown here]

W. Shi, E. Collins, and V. Karamcheti. Modeling object characteristics of dynamic web content. IEEE Globecomm 2002.

....views of the My Yahoo portal, different users end up sharing the same news headlines and TV program guides. Two recent studies have shown that approximately 60 of the bytes in dynamic responses from a set of popular web sites could in fact be reused from a previous retrieval of the page [34, 39]. Although encouraging, the above proposals and studies need to be supplemented with characterizations of the actual workload encountered on sites that serve dynamic and personalized content. These characterizations serve two roles: first, they provide evidence for whether or not object ....

....and 99 are less than 3000 bytes. We found the channel sizes are best modeled using a Weibull distribution (with CDF F(X) 1 e (0. 00m) This observation is in agreement with a previous study where we looked at object sizes in dynamic documents downloaded from six news and e commerce web sites [34]. It is interesting to compare this distribution with the overall document size distribution in Figure 7(b) The latter shows that 70 of the documents lie in a very small range between 9,725 and 10,688 bytes. The popularity of the HONE tab and the fact that the layout accounts for a sizeable ....

[Article contains additional citation context not shown here]

W. Shi, E. Collins, and V. Karamcheti. Modeling object characteristics of dynamic web content. Tech. Rep. TR2001.

....size exceeds a certain threshold (and whose child nodes have size smaller than the threshold) Level based splitting follows the logical structure of the document: all nodes below a certain depth in the tree are grouped together. More details of the methodology can be found in a technical report [22]. Using this methodology, we analyzed traces collected from three news sites (www. cnn. cam, dailynews. yahaa. cam, www.nytimes.cam) tWO e commerce sites (www. amazon. cam, www. barnesnoble. cam) and an entertainment site (www. windowsmedia. cam) The main pages at these sites were downloaded ....

W. Shi, E. Collins, and V. Karamcheti. Modeling object characteristics of dynamic web content. Tech. Rep. TR2001.

....size exceeds a certain threshold (and whose child nodes have size smaller than the threshold) Level based splitting follows the logical structure of the document: all nodes below a certain depth in the tree are grouped together. More details of the methodology can be found in a technical report [23]. Using this methodology, we analyzed traces collected from three news sites (www.cnn.com, dailynews.yahoo.com, www.nytimes.com) two e commerce sites (www.amazon. com, www.barnesnoble.com) and an entertainment site (www.windowsmedia.com) The main pages at these sites were downloaded every ten ....

W. Shi, E. Collins, and V. Karamcheti. Modeling object characteristics of dynamic web content. Tech. Rep. TR2001-822, Computer Science Department, New York University, Nov. 2001, http://www. cs.nyu.edu/weisong/papers/tr2001-822.pdf.

....at which document characteristics are modeled. To cope with the absence of an explicit template in the documents, we inferred both the template and the component objects using parameterized level based and size based splitting techniques described in additional detail in a technical report [10] 2.1 Analysis of Dynamic Content Characteristics We analyzed traces collected from six representative Web sites, with frequently changing dynamic content: three news sites (www.cnn. com, dailynews.yahoo.com, www.nytimes.com) two e commerce sites (www.amazon.com, www.barnesnoble.com) and an ....

....objects) a) size (b) freshness time Figure 1: The measured cumulative distribution of (a) object sizes and (b) freshness times for different size limit settings settings for traces collected from the www.cnn.com site. Each of these traces was analyzed using the methodology described earlier [10]. Due to space restrictions, we discuss only the results for the cnn trace, which is representative of the others. Figure 1 shows the measured cumulative distribution of object sizes and freshness times for different size limit settings; the latter parameter denotes the target of the document ....

W. Shi, E. Collins, and V. Karamcheti. Modeling object characteristics of dynamic web content. Tech. Rep. TR2001-822, Computer Science Department, New York University, Nov. 2001, http://www.cs.nyu.edu/weisong/papers/tr2001-822.pdf.

.... for reuse at the subdocument level (at the granularity of individual objects making up the overall document) Two recent studies have shown that approximately 60 of the bytes in dynamic responses from a set of popular web sites could in fact be reused from a previous retrieval of the page [34, 39]. Although encouraging, the above proposals and studies need to be supplemented with characterizations of the actual workload encountered on sites that serve dynamic and personalized content. These characterizations serve two roles: first, they provide evidence for whether or not object ....

....of the distribution of channel size. We find that the channel sizes, 99 of which are smaller than 3000 bytes, are best modeled using a Weibull distribution (with CDF F (X) 1 e (0.0012x) 1. 6 ) This observation is in agreement with a previous study on six news and e commerce web sites [34]. It is interesting to compare this distribution with the overall document size distribution in Figure 7(b) The latter shows that 70 of the documents lie in a very small range between 9,725 and 10,688 bytes. The popularity of the HOME tab and the fact that the template accounts for a sizeable ....

W. Shi, E. Collins, and V. Karamcheti. Modeling object characteristics of dynamic web content. IEEE Globecomm 2002.

....at which document characteristics are modeled. To cope with the absence of an explicit template in the documents, we inferred both the template and the component objects using parameterized level based and size based splitting techniques described in additional detail in a technical report [ 10] 2.1 Analysis of Dynamic Content Characteristics We analyzed traces collected from six representative Web sites, with frequently changing dynamic content: three news sites (www. cnn. cam, dailynews. yahaa. cam, www. nytimes. cam) two e commerce sites (www. amazon. cam, www. barnesnoble. cam) ....

....news sites (www. cnn. cam, dailynews. yahaa. cam, www. nytimes. cam) two e commerce sites (www. amazon. cam, www. barnesnoble. cam) and an entertainment site (www. windowsmedia. cam) The traces consisted of downloads of the main page at each of the sites, every ten minutes over a two week interval. 1.0 0.9 0.8 0.7 sizelimit=O 0.4 sizelimit=500 . sizelimit=1000 0.3 100 1000 10000 Objectsize 100 200 300 400 500 600 700 800 900 100011001200130014001500 Time(Minutes) a) size (b) freshness time Figure 1: The measured cumulative distribution of (a) object sizes ....

[Article contains additional citation context not shown here]

W. Shi, E. Collins, and V. Karamcheti. Modeling object characteristics of dynamic web content. Tech. Rep. TR200 b822, Computer Science Depamnent, New York University, Nov. 200t, tTttp://www.cs.nyu.edu/-weisong/papers/tr2001 822.pdf.

No context found.

Weisong Shi, Eli Collins, and Vijay Karamcheti. Modeling Object Characteristics of Dynamic Web Content. Special Issue on scalable Internet services and architecture of Journal of Parallel and Distributed Computing (JPDC), Sept. 2003.

No context found.

W. Shi, E. Collins, and V. Karamcheti. Modeling Object Characteristics of Dynamic Web Content. Special Issue on scalable Internet services and architecture of Journal of Parallel and Distributed Computing (JPDC), Sept. 2003.

No context found.

W. Shi, R. Wright, E. Collins, and V. Karamcheti, "Modeling object characteristics of dynamic web content," Journal of Par. and Distributed Computing, 2003.

No context found.

Shi, W., Wright, R., Collins, E. and Karamcheti, V. (2003) `Modeling object characteristics of dynamic web content', Journal of Parallel and Distributed Computing (JPDC), special issue on scalable internet services and architecture.

No context found.

W. Shi, E. Collins, and V. Karamcheti. Modeling object characteristics of dynamic web content. Journal of Parallel and Distributed Computing (to appear), Sept. 2003.

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