2.2
Paper: (7) World Wide Web Cache Consistency
Positive Aspects of the Paper:
*A comparision between the TTL style of web caching and client polling (Alex caching) is taken over several factors , such as web hierarchy , Last Modified Page value, page freshness etc.
*Further access patterns of data is taken into account during the optimizing discussion. Access patterns of data are key factors to any caching protocol.A great deal of invalidation protocols being reduced in cases where none is required can reduce bandwidth consumption and server load greatly. It is here that the Alex protocol beats the TTL style hands down. An update threshold factor taken into account for the Alex caching while at the same time considering access patterns displays the strength of the same.
*A real world study is undertaken by comparing two scenarios- a business perspective using the Microsoft proxy cache and a research perspective using the Boston University web server. Both studies being compared over a large sample of time , and equally offer justice to the comparison.
*Finally it is shown that in certain cases where access patterns cannot be determined, the TTL style would be the best bet. This is obvious in the case where object lifetimes are known 'a priori'. It is obvious that in such a case, you would only let the oject live until it expires and finally send the new object in place of it. Hence redundant invalidation protocols are avoided and access patterns are constant for the object.
Negative Aspects of the Paper:
*A third form of caching is ignored largely- the ivalidation protocols technique where servers co-operate among each other at storing data and invalidating cached copies is a common study of distributed systems. The web can be viewed as a large distributed system- the paper also mentions that modern day servers cache data at multiple locations. This is common of distributed systems as well. A co-ordinated invalidation protocol achieve can optimal performance. This aspect of caching is mentioned at the start of the paper , but ignored at large later.
*During the optimization discussion, the major chunk is absorbed by access patterns optimization techniques. This is certainly necessary and must be taken into account- however most research indicates a change in perspective of the current protocol, and sometimes suggest revamping the protocol entirely to suit the current day needs. This paper hints no such change. While protocols can be optimized, exponential growth of data generated and consumed will soon exhaust the optimized solution as well. It would be imperative for the research community to suggest changes at large and at times revamp the entire protocol.
*The sample of study is taken over a fairly large amount of time- this is necessary. However it would be necessary to even determine access patterns during the different times of the day. At times protocols which perform averagely over a certain period of time indicate that they perform consistently at all times of the day. Since the web is a large system spanning over different networks, the study should have included the ...