The Anatomy of a Large Scale Hypertextual Web Search Engine
Lawrence Page, Sergey Brin

Problems

This paper discusses Google, a new web search engine that aims to solve many
emerging challenges of finding information on the rapidly exploding world wide
web.

(1) Scaling: Google is designed with the goal of being able to scale and work
as the web grows in size (amount of information). Scalability becomes essential
as the web grows at breakneck speed. Storage space needs to be used 
efficiently, the indexing system must process hundreds of gigabytes of data
efficiently and the search engine must be able to handle millions of queries
every day (thousands of queries a second).

(2) Quality of Search: Despite the information explosion on the web, search 
engine users are typically interested only in the first few links returned
by search engine. Given this constraint, the number of matches returned by
an engine takes a back seat to the relevance and quality of the first few links
returned by the search engine since those are the links that will be used by
the user. Google attempts to provide better quality search results by 
introducing a novel ranking(PageRank) system that is used to prioritize the results.

Strengths and New Ideas

(1) PageRank: The PageRank is the striking feature of the google search engine.
Google's emphasis on better quality results than on indexing a larger portion
of the web is justified by the fact that a typical user rarely looks beyond
the first few matches returned by the search engine. Google has created citation
graphs that allow them to calculate a web page's "Page Rank". Page Rank is a 
measure of a web page's citation importance that aligns well with the human 
idea of importance-hence page rank becomes an effective measure to 
prioritize the search results. Page rank is based on the idea that a page's 
importance goes up based on the number of back links it has--intuitively, the
more the number of back links, the more useful the page, since people would tend
to link to "good quality" information and a lot of links to one page is a strong
indicator of people agreeing to the good quality of information available on 
that page. Besides, the rank of the page also depends on the page rank of 
the pages that point to it. A better quality page would intuitively point to
good information. 

(2) Anchor Text

Google also uses the text of links to refer to pages that the link points to
in addition to the page that the link is on. This is based on the fact that
link text is often a more accurate descriptor of web pages than the information
on the pages itself. In addition, the association of links to the pages they
point to allows google to index a wider variety of documents such as images, 
programs, databases etc.

(3) Mention of Bottlenecks

The paper also discusses the bottlenecks that were encountered in the 
development of google the main one being disk seek time. While Google's focus
was not on efficiency, for a commercial search engine requiring to handl
millions of queries, fast response time is crucial. The description of the 
bottlenecks alerts people new to the field to possible pitfalls to be careful
of while designing their own search engines. 

Weaknesses and Extensions

(1) Although the authors have discussed the data structures and major 
components, it is not thorough and detailed. For a system as big and
complicated as Google, there has to be a lot more going on than is covered
in the paper. The discussion presented in the paper is more an overview.

(2) The use of anchor links might have a possible negative aspect to it. The
authors agree that achor links might point to non-existent pages although
they reduce the probability of such an occurence by sorting the results. While
the number of anchor links returned can be handled currently, will it be
able to scale when the web grows by an order of magnitude. This would be 
something that might need to be looked into.

(3) The authors claim that Google provides better results than other commercial
search engines. However, a detailed comparison between Google and other search
engines is not provided.