History page-brin thesis - anatomy of a large scale hypertextual web search engine

inf olab.st anf ord.edu http://infolab.stanford.edu/~backrub/google.html
The Anatomy of a Large-Scale Hypertextual Web Search Engine
Sergey Brin and Lawrence Page
{sergey, page}@cs.stanf ord.edu
Computer Science Department, Stanf ord University, Stanf ord, CA 94305
Abstract
In this paper, we present Google, a prototype of a large-scale search engine which makes
heavy use of the structure present in hypertext. Google is designed to crawl and index the Web
efficiently and produce much more satisfying search results than existing systems. The prototype
with a full text and hyperlink database of at least 24 million pages is available at
http://google.stanford.edu/
To engineer a search engine is a challenging task. Search engines index tens to hundreds of
millions of web pages involving a comparable number of distinct terms. They answer tens of millions
of queries every day. Despite the importance of large-scale search engines on the web, very little
academic research has been done on them. Furthermore, due to rapid advance in technology and
web proliferation, creating a web search engine today is very different from three years ago. This
paper provides an in-depth description of our large-scale web search engine -- the first such
detailed public description we know of to date.
Apart from the problems of scaling traditional search techniques to data of this magnitude, there are
new technical challenges involved with using the additional information present in hypertext to
produce better search results. This paper addresses this question of how to build a practical large-
scale system which can exploit the additional information present in hypertext. Also we look at the
problem of how to effectively deal with uncontrolled hypertext collections where anyone can publish
anything they want.
Keywords: World Wide Web, Search Engines, Information Retrieval, PageRank, Google
1. Introduction
(Note: There are two versions of this paper -- a longer full version and a shorter printed version. The full version is
available on the web and the conference CD-ROM.) The web creates new challenges f or inf ormation retrieval.
The amount of inf ormation on the web is growing rapidly, as well as the number of new users inexperienced in
the art of web research. People are likely to surf the web using its link graph, of ten starting with high quality
human maintained indices such as Yahoo! or with search engines. Human maintained lists cover popular topics
ef f ectively but are subjective, expensive to build and maintain, slow to improve, and cannot cover all esoteric
topics. Automated search engines that rely on keyword matching usually return too many low quality matches.
To make matters worse, some advertisers attempt to gain people's attention by taking measures meant to
mislead automated search engines. We have built a large-scale search engine which addresses many of the
problems of existing systems. It makes especially heavy use of the additional structure present in hypertext to
provide much higher quality search results. We chose our system name, Google, because it is a common
spelling of googol, or 10100 and f its well with our goal of building very large-scale search engines.
1.1 Web Search Engines -- Scaling Up: 1994 - 2000

Search engine technology has had to scale dramatically to keep up with the growth of the web. In 1994, one of
the f irst web search engines, the World Wide Web Worm (WWWW) [McBryan 94] had an index of 110,000 web
pages and web accessible documents. As of November, 1997, the top search engines claim to index f rom 2
million (WebCrawler) to 100 million web documents (f rom Search Engine Watch). It is f oreseeable that by the
year 2000, a comprehensive index of the Web will contain over a billion documents. At the same time, the
number of queries search engines handle has grown incredibly too. In March and April 1994, the World Wide
Web Worm received an average of about 1500 queries per day. In November 1997, Altavista claimed it handled
roughly 20 million queries per day. With the increasing number of users on the web, and automated systems
which query search engines, it is likely that top search engines will handle hundreds of millions of queries per
day by the year 2000. The goal of our system is to address many of the problems, both in quality and
scalability, introduced by scaling search engine technology to such extraordinary numbers.
1.2. Google: Scaling with the Web
Creating a search engine which scales even to today's web presents many challenges. Fast crawling
technology is needed to gather the web documents and keep them up to date. Storage space must be used
ef f iciently to store indices and, optionally, the documents themselves. The indexing system must process
hundreds of gigabytes of data ef f iciently. Queries must be handled quickly, at a rate of hundreds to thousands
per second.
These tasks are becoming increasingly dif f icult as the Web grows. However, hardware perf ormance and cost
have improved dramatically to partially of f set the dif f iculty. There are, however, several notable exceptions to
this progress such as disk seek time and operating system robustness. In designing Google, we have
considered both the rate of growth of the Web and technological changes. Google is designed to scale well to
extremely large data sets. It makes ef f icient use of storage space to store the index. Its data structures are
optimized f or f ast and ef f icient access (see section 4.2). Further, we expect that the cost to index and store
text or HTML will eventually decline relative to the amount that will be available (see Appendix B). This will result
in f avorable scaling properties f or centralized systems like Google.
1.3 Design Goals
1.3.1 Improved Search Quality
Our main goal is to improve the quality of web search engines. In 1994, some people believed that a complete
search index would make it possible to f ind anything easily. According to Best of the Web 1994 -- Navigators,
"The best navigation service should make it easy to f ind almost anything on the Web (once all the data is
entered)." However, the Web of 1997 is quite dif f erent. Anyone who has used a search engine recently, can
readily testif y that the completeness of the index is not the only f actor in the quality of search results. "Junk
results" of ten wash out any results that a user is interested in. In f act, as of November 1997, only one of the
top f our commercial search engines f inds itself (returns its own search page in response to its name in the top
ten results). One of the main causes of this problem is that the number of documents in the indices has been
increasing by many orders of magnitude, but the user's ability to look at documents has not. People are still
only willing to look at the f irst f ew tens of results. Because of this, as the collection size grows, we need tools
that have very high precision (number of relevant documents returned, say in the top tens of results). Indeed,
we want our notion of "relevant" to only include the very best documents since there may be tens of
thousands of slightly relevant documents. This very high precision is important even at the expense of recall
(the total number of relevant documents the system is able to return). There is quite a bit of recent optimism
that the use of more hypertextual inf ormation can help improve search and other applications [Marchiori 97]
[Spertus 97] [Weiss 96] [Kleinberg 98]. In particular, link structure [Page 98] and link text provide a lot of
inf ormation f or making relevance judgments and quality f iltering. Google makes use of both link structure and

anchor text (see Sections 2.1 and 2.2).
1.3.2 Academic Search Engine Research
Aside f rom tremendous growth, the Web has also become increasingly commercial over time. In 1993, 1.5% of
web servers were on .com domains. This number grew to over 60% in 1997. At the same time, search engines
have migrated f rom the academic domain to the commercial. Up until now most search engine development has
gone on at companies with little publication of technical details. This causes search engine technology to
remain largely a black art and to be advertising oriented (see Appendix A). With Google, we have a strong goal
to push more development and understanding into the academic realm.
Another important design goal was to build systems that reasonable numbers of people can actually use.
Usage was important to us because we think some of the most interesting research will involve leveraging the
vast amount of usage data that is available f rom modern web systems. For example, there are many tens of
millions of searches perf ormed every day. However, it is very dif f icult to get this data, mainly because it is
considered commercially valuable.
Our f inal design goal was to build an architecture that can support novel research activities on large-scale web
data. To support novel research uses, Google stores all of the actual documents it crawls in compressed f orm.
One of our main goals in designing Google was to set up an environment where other researchers can come in
quickly, process large chunks of the web, and produce interesting results that would have been very dif f icult to
produce otherwise. In the short time the system has been up, there have already been several papers using
databases generated by Google, and many others are underway. Another goal we have is to set up a Spacelab-
like environment where researchers or even students can propose and do interesting experiments on our
large-scale web data.
2. System Features
The Google search engine has two important f eatures that help it produce high precision results. First, it
makes use of the link structure of the Web to calculate a quality ranking f or each web page. This ranking is
called PageRank and is described in detail in [Page 98]. Second, Google utilizes link to improve search results.
2.1 PageRank: Bringing Order to the Web
The citation (link) graph of the web is an important resource that has largely gone unused in existing web
search engines. We have created maps containing as many as 518 million of these hyperlinks, a signif icant
sample of the total. These maps allow rapid calculation of a web page's "PageRank", an objective measure of
its citation importance that corresponds well with people's subjective idea of importance. Because of this
correspondence, PageRank is an excellent way to prioritize the results of web keyword searches. For most
popular subjects, a simple text matching search that is restricted to web page titles perf orms admirably when
PageRank prioritizes the results (demo available at google.stanf ord.edu). For the type of f ull text searches in
the main Google system, PageRank also helps a great deal.
2.1.1 Description of PageRank Calculation
Academic citation literature has been applied to the web, largely by counting citations or backlinks to a given
page. This gives some approximation of a page's importance or quality. PageRank extends this idea by not
counting links f rom all pages equally, and by normalizing by the number of links on a page. PageRank is def ined
as f ollows:

We assume page A has pages T1...Tn which point to it (i.e., are citations). The parameter d is a
damping factor which can be set between 0 and 1. We usually set d to 0.85. There are more details
about d in the next section. Also C(A) is defined as the number of links going out of page A. The
PageRank of a page A is given as follows:
PR(A) = (1-d) + d (PR(T1)/C(T1) + ... + PR(Tn)/C(Tn))
Note that the PageRanks form a probability distribution over web pages, so the sum of all web
pages' PageRanks will be one.
PageRank or PR(A) can be calculated using a simple iterative algorithm, and corresponds to the principal
eigenvector of the normalized link matrix of the web. Also, a PageRank f or 26 million web pages can be
computed in a f ew hours on a medium size workstation. There are many other details which are beyond the
scope of this paper.
2.1.2 Intuitive Justification
PageRank can be thought of as a model of user behavior. We assume there is a "random surf er" who is given
a web page at random and keeps clicking on links, never hitting "back" but eventually gets bored and starts on
another random page. The probability that the random surf er visits a page is its PageRank. And, the d damping
f actor is the probability at each page the "random surf er" will get bored and request another random page. One
important variation is to only add the damping f actor d to a single page, or a group of pages. This allows f or
personalization and can make it nearly impossible to deliberately mislead the system in order to get a higher
ranking. We have several other extensions to PageRank, again see [Page 98].
Another intuitive justif ication is that a page can have a high PageRank if there are many pages that point to it,
or if there are some pages that point to it and have a high PageRank. Intuitively, pages that are well cited f rom
many places around the web are worth looking at. Also, pages that have perhaps only one citation f rom
something like the Yahoo! homepage are also generally worth looking at. If a page was not high quality, or was
a broken link, it is quite likely that Yahoo's homepage would not link to it. PageRank handles both these cases
and everything in between by recursively propagating weights through the link structure of the web.
2.2 Anchor Text
The text of links is treated in a special way in our search engine. Most search engines associate the text of a
link with the page that the link is on. In addition, we associate it with the page the link points to. This has
several advantages. First, anchors of ten provide more accurate descriptions of web pages than the pages
themselves. Second, anchors may exist f or documents which cannot be indexed by a text-based search
engine, such as images, programs, and databases. This makes it possible to return web pages which have not
actually been crawled. Note that pages that have not been crawled can cause problems, since they are never
checked f or validity bef ore being returned to the user. In this case, the search engine can even return a page
that never actually existed, but had hyperlinks pointing to it. However, it is possible to sort the results, so that
this particular problem rarely happens.
This idea of propagating anchor text to the page it ref ers to was implemented in the World Wide Web Worm
[McBryan 94] especially because it helps search non-text inf ormation, and expands the search coverage with
f ewer downloaded documents. We use anchor propagation mostly because anchor text can help provide better
quality results. Using anchor text ef f iciently is technically dif f icult because of the large amounts of data which
must be processed. In our current crawl of 24 million pages, we had over 259 million anchors which we indexed.

2.3 Other Features
Aside f rom PageRank and the use of anchor text, Google has several other f eatures. First, it has location
inf ormation f or all hits and so it makes extensive use of proximity in search. Second, Google keeps track of
some visual presentation details such as f ont size of words. Words in a larger or bolder f ont are weighted
higher than other words. Third, f ull raw HTML of pages is available in a repository.
3 Related Work
Search research on the web has a short and concise history. The World Wide Web Worm (WWWW) [McBryan
94] was one of the f irst web search engines. It was subsequently f ollowed by several other academic search
engines, many of which are now public companies. Compared to the growth of the Web and the importance of
search engines there are precious f ew documents about recent search engines [Pinkerton 94]. According to
Michael Mauldin (chief scientist, Lycos Inc) [Mauldin], "the various services (including Lycos) closely guard the
details of these databases". However, there has been a f air amount of work on specif ic f eatures of search
engines. Especially well represented is work which can get results by post-processing the results of existing
commercial search engines, or produce small scale "individualized" search engines. Finally, there has been a lot
of research on inf ormation retrieval systems, especially on well controlled collections. In the next two sections,
we discuss some areas where this research needs to be extended to work better on the web.
3.1 Information Retrieval
Work in inf ormation retrieval systems goes back many years and is well developed [Witten 94]. However, most
of the research on inf ormation retrieval systems is on small well controlled homogeneous collections such as
collections of scientif ic papers or news stories on a related topic. Indeed, the primary benchmark f or
inf ormation retrieval, the Text Retrieval Conf erence [TREC 96], uses a f airly small, well controlled collection f or
their benchmarks. The "Very Large Corpus" benchmark is only 20GB compared to the 147GB f rom our crawl of
24 million web pages. Things that work well on TREC of ten do not produce good results on the web. For
example, the standard vector space model tries to return the document that most closely approximates the
query, given that both query and document are vectors def ined by their word occurrence. On the web, this
strategy of ten returns very short documents that are the query plus a f ew words. For example, we have seen a
major search engine return a page containing only "Bill Clinton Sucks" and picture f rom a "Bill Clinton" query.
Some argue that on the web, users should specif y more accurately what they want and add more words to
their query. We disagree vehemently with this position. If a user issues a query like "Bill Clinton" they should get
reasonable results since there is a enormous amount of high quality inf ormation available on this topic. Given
examples like these, we believe that the standard inf ormation retrieval work needs to be extended to deal
ef f ectively with the web.
3.2 Differences Between the Web and Well Controlled Collections
The web is a vast collection of completely uncontrolled heterogeneous documents. Documents on the web
have extreme variation internal to the documents, and also in the external meta inf ormation that might be
available. For example, documents dif f er internally in their language (both human and programming), vocabulary
(email addresses, links, zip codes, phone numbers, product numbers), type or f ormat (text, HTML, PDF,
images, sounds), and may even be machine generated (log f iles or output f rom a database). On the other
hand, we def ine external meta inf ormation as inf ormation that can be inf erred about a document, but is not
contained within it. Examples of external meta inf ormation include things like reputation of the source, update
f requency, quality, popularity or usage, and citations. Not only are the possible sources of external meta
inf ormation varied, but the things that are being measured vary many orders of magnitude as well. For example,
compare the usage inf ormation f rom a major homepage, like Yahoo's which currently receives millions of page
views every day with an obscure historical article which might receive one view every ten years. Clearly, these

Figure 1. High Level Google Architecture
two items must be treated very dif f erently by a search engine.
Another big dif f erence between the web and traditional well controlled collections is that there is virtually no
control over what people can put on the web. Couple this f lexibility to publish anything with the enormous
inf luence of search engines to route traf f ic and companies which deliberately manipulating search engines f or
prof it become a serious problem. This problem that has not been addressed in traditional closed inf ormation
retrieval systems. Also, it is interesting to note that metadata ef f orts have largely f ailed with web search
engines, because any text on the page which is not directly represented to the user is abused to manipulate
search engines. There are even numerous companies which specialize in manipulating search engines f or
prof it.
4 System Anatomy
First, we will provide a high level discussion of the architecture. Then, there is some in-depth descriptions of
important data structures. Finally, the major applications: crawling, indexing, and searching will be examined in
depth.
4.1 Google Architecture Overview
In this section, we will give a high level overview of how the
whole system works as pictured in Figure 1. Further
sections will discuss the applications and data structures
not mentioned in this section. Most of Google is
implemented in C or C++ f or ef f iciency and can run in either
Solaris or Linux.
In Google, the web crawling (downloading of web pages) is
done by several distributed crawlers. There is a URLserver
that sends lists of URLs to be f etched to the crawlers. The
web pages that are f etched are then sent to the
storeserver. The storeserver then compresses and stores
the web pages into a repository. Every web page has an
associated ID number called a docID which is assigned
whenever a new URL is parsed out of a web page. The
indexing f unction is perf ormed by the indexer and the
sorter. The indexer perf orms a number of f unctions. It
reads the repository, uncompresses the documents, and
parses them. Each document is converted into a set of
word occurrences called hits. The hits record the word,
position in document, an approximation of f ont size, and capitalization. The indexer distributes these hits into
a set of "barrels", creating a partially sorted f orward index. The indexer perf orms another important f unction. It
parses out all the links in every web page and stores important inf ormation about them in an anchors f ile. This
f ile contains enough inf ormation to determine where each link points f rom and to, and the text of the link.
The URLresolver reads the anchors f ile and converts relative URLs into absolute URLs and in turn into docIDs.
It puts the anchor text into the f orward index, associated with the docID that the anchor points to. It also
generates a database of links which are pairs of docIDs. The links database is used to compute PageRanks
f or all the documents.
The sorter takes the barrels, which are sorted by docID (this is a simplif ication, see Section 4.2.5), and resorts
them by wordID to generate the inverted index. This is done in place so that little temporary space is needed

Figure 2. Repository Data Structure
f or this operation. The sorter also produces a list of wordIDs and of f sets into the inverted index. A program
called DumpLexicon takes this list together with the lexicon produced by the indexer and generates a new
lexicon to be used by the searcher. The searcher is run by a web server and uses the lexicon built by
DumpLexicon together with the inverted index and the PageRanks to answer queries.
4.2 Major Data Structures
Google's data structures are optimized so that a large document collection can be crawled, indexed, and
searched with little cost. Although, CPUs and bulk input output rates have improved dramatically over the years,
a disk seek still requires about 10 ms to complete. Google is designed to avoid disk seeks whenever possible,
and this has had a considerable inf luence on the design of the data structures.
4.2.1 BigFiles
BigFiles are virtual f iles spanning multiple f ile systems and are addressable by 64 bit integers. The allocation
among multiple f ile systems is handled automatically. The BigFiles package also handles allocation and
deallocation of f ile descriptors, since the operating systems do not provide enough f or our needs. BigFiles
also support rudimentary compression options.
4.2.2 Repository
The repository contains the f ull HTML of every web page.
Each page is compressed using zlib (see RFC1950). The
choice of compression technique is a tradeof f between speed
and compression ratio. We chose zlib's speed over a
signif icant improvement in compression of f ered by bzip. The
compression rate of bzip was approximately 4 to 1 on the
repository as compared to zlib's 3 to 1 compression. In the
repository, the documents are stored one af ter the other and
are pref ixed by docID, length, and URL as can be seen in
Figure 2. The repository requires no other data structures to
be used in order to access it. This helps with data
consistency and makes development much easier; we can rebuild all the other data structures f rom only the
repository and a f ile which lists crawler errors.
4.2.3 Document Index
The document index keeps inf ormation about each document. It is a f ixed width ISAM (Index sequential access
mode) index, ordered by docID. The inf ormation stored in each entry includes the current document status, a
pointer into the repository, a document checksum, and various statistics. If the document has been crawled, it
also contains a pointer into a variable width f ile called docinf o which contains its URL and title. Otherwise the
pointer points into the URLlist which contains just the URL. This design decision was driven by the desire to
have a reasonably compact data structure, and the ability to f etch a record in one disk seek during a search
Additionally, there is a f ile which is used to convert URLs into docIDs. It is a list of URL checksums with their
corresponding docIDs and is sorted by checksum. In order to f ind the docID of a particular URL, the URL's
checksum is computed and a binary search is perf ormed on the checksums f ile to f ind its docID. URLs may be
converted into docIDs in batch by doing a merge with this f ile. This is the technique the URLresolver uses to
turn URLs into docIDs. This batch mode of update is crucial because otherwise we must perf orm one seek f or

Figure
3.
every link which assuming one disk would take more than a month f or our 322 million link dataset.
4.2.4 Lexicon
The lexicon has several dif f erent f orms. One important change f rom earlier systems is that the lexicon can f it
in memory f or a reasonable price. In the current implementation we can keep the lexicon in memory on a
machine with 256 MB of main memory. The current lexicon contains 14 million words (though some rare words
were not added to the lexicon). It is implemented in two parts -- a list of the words (concatenated together but
separated by nulls) and a hash table of pointers. For various f unctions, the list of words has some auxiliary
inf ormation which is beyond the scope of this paper to explain f ully.
4.2.5 Hit Lists
A hit list corresponds to a list of occurrences of a particular word in a particular document including position,
f ont, and capitalization inf ormation. Hit lists account f or most of the space used in both the f orward and the
inverted indices. Because of this, it is important to represent them as ef f iciently as possible. We considered
several alternatives f or encoding position, f ont, and capitalization -- simple encoding (a triple of integers), a
compact encoding (a hand optimized allocation of bits), and Huf f man coding. In the end we chose a hand
optimized compact encoding since it required f ar less space than the simple encoding and f ar less bit
manipulation than Huf f man coding. The details of the hits are shown in Figure 3.
Our compact encoding uses two bytes f or every hit. There are two types of hits: f ancy hits and plain hits.
Fancy hits include hits occurring in a URL, title, anchor text, or meta tag. Plain hits include everything else. A
plain hit consists of a capitalization bit, f ont size, and 12 bits of word position in a document (all positions
higher than 4095 are labeled 4096). Font size is represented relative to the rest of the document using three
bits (only 7 values are actually used because 111 is the f lag that signals a f ancy hit). A f ancy hit consists of a
capitalization bit, the f ont size set to 7 to indicate it is a f ancy hit, 4 bits to encode the type of f ancy hit, and 8
bits of position. For anchor hits, the 8 bits of position are split into 4 bits f or position in anchor and 4 bits f or
a hash of the docID the anchor occurs in. This gives us some limited phrase searching as long as there are not
that many anchors f or a particular word. We expect to update the way that anchor hits are stored to allow f or
greater resolution in the position and docIDhash f ields. We use f ont size relative to the rest of the document
because when searching, you do not want to rank otherwise identical documents dif f erently just because one
of the documents is in a larger f ont.
The length of a hit list is stored bef ore the hits
themselves. To save space, the length of the hit list is
combined with the wordID in the f orward index and
the docID in the inverted index. This limits it to 8 and 5
bits respectively (there are some tricks which allow 8
bits to be borrowed f rom the wordID). If the length is
longer than would f it in that many bits, an escape
code is used in those bits, and the next two bytes
contain the actual length.
4.2.6 Forward Index
The f orward index is actually already partially sorted.
It is stored in a number of barrels (we used 64). Each

Forward and Reverse Indexes and the Lexicon
barrel holds a range of wordID's. If a document
contains words that f all into a particular barrel, the
docID is recorded into the barrel, f ollowed by a list of
wordID's with hitlists which correspond to those
words. This scheme requires slightly more storage
because of duplicated docIDs but the dif f erence is very small f or a reasonable number of buckets and saves
considerable time and coding complexity in the f inal indexing phase done by the sorter. Furthermore, instead of
storing actual wordID's, we store each wordID as a relative dif f erence f rom the minimum wordID that f alls into
the barrel the wordID is in. This way, we can use just 24 bits f or the wordID's in the unsorted barrels, leaving 8
bits f or the hit list length.
4.2.7 Inverted Index
The inverted index consists of the same barrels as the f orward index, except that they have been processed
by the sorter. For every valid wordID, the lexicon contains a pointer into the barrel that wordID f alls into. It
points to a doclist of docID's together with their corresponding hit lists. This doclist represents all the
occurrences of that word in all documents.
An important issue is in what order the docID's should appear in the doclist. One simple solution is to store
them sorted by docID. This allows f or quick merging of dif f erent doclists f or multiple word queries. Another
option is to store them sorted by a ranking of the occurrence of the word in each document. This makes
answering one word queries trivial and makes it likely that the answers to multiple word queries are near the
start. However, merging is much more dif f icult. Also, this makes development much more dif f icult in that a
change to the ranking f unction requires a rebuild of the index. We chose a compromise between these options,
keeping two sets of inverted barrels -- one set f or hit lists which include title or anchor hits and another set
f or all hit lists. This way, we check the f irst set of barrels f irst and if there are not enough matches within
those barrels we check the larger ones.
4.3 Crawling the Web
Running a web crawler is a challenging task. There are tricky perf ormance and reliability issues and even more
importantly, there are social issues. Crawling is the most f ragile application since it involves interacting with
hundreds of thousands of web servers and various name servers which are all beyond the control of the
system.
In order to scale to hundreds of millions of web pages, Google has a f ast distributed crawling system. A single
URLserver serves lists of URLs to a number of crawlers (we typically ran about 3). Both the URLserver and the
crawlers are implemented in Python. Each crawler keeps roughly 300 connections open at once. This is
necessary to retrieve web pages at a f ast enough pace. At peak speeds, the system can crawl over 100 web
pages per second using f our crawlers. This amounts to roughly 600K per second of data. A major perf ormance
stress is DNS lookup. Each crawler maintains a its own DNS cache so it does not need to do a DNS lookup
bef ore crawling each document. Each of the hundreds of connections can be in a number of dif f erent states:
looking up DNS, connecting to host, sending request, and receiving response. These f actors make the crawler
a complex component of the system. It uses asynchronous IO to manage events, and a number of queues to
move page f etches f rom state to state.
It turns out that running a crawler which connects to more than half a million servers, and generates tens of
millions of log entries generates a f air amount of email and phone calls. Because of the vast number of people
coming on line, there are always those who do not know what a crawler is, because this is the f irst one they
have seen. Almost daily, we receive an email something like, "Wow, you looked at a lot of pages f rom my web
site. How did you like it?" There are also some people who do not know about the robots exclusion protocol,

1. Parse the query.
2. Convert words into wordIDs.
3. Seek to the start of the doclist in the short barrel f or every word.
and think their page should be protected f rom indexing by a statement like, "This page is copyrighted and
should not be indexed", which needless to say is dif f icult f or web crawlers to understand. Also, because of the
huge amount of data involved, unexpected things will happen. For example, our system tried to crawl an online
game. This resulted in lots of garbage messages in the middle of their game! It turns out this was an easy
problem to f ix. But this problem had not come up until we had downloaded tens of millions of pages. Because
of the immense variation in web pages and servers, it is virtually impossible to test a crawler without running it
on large part of the Internet. Invariably, there are hundreds of obscure problems which may only occur on one
page out of the whole web and cause the crawler to crash, or worse, cause unpredictable or incorrect behavior.
Systems which access large parts of the Internet need to be designed to be very robust and caref ully tested.
Since large complex systems such as crawlers will invariably cause problems, there needs to be signif icant
resources devoted to reading the email and solving these problems as they come up.
4.4 Indexing the Web
Parsing -- Any parser which is designed to run on the entire Web must handle a huge array of possible
errors. These range f rom typos in HTML tags to kilobytes of zeros in the middle of a tag, non-ASCII
characters, HTML tags nested hundreds deep, and a great variety of other errors that challenge
anyone's imagination to come up with equally creative ones. For maximum speed, instead of using YACC
to generate a CFG parser, we use f lex to generate a lexical analyzer which we outf it with its own stack.
Developing this parser which runs at a reasonable speed and is very robust involved a f air amount of
work.
Indexing Documents into Barrels -- Af ter each document is parsed, it is encoded into a number of
barrels. Every word is converted into a wordID by using an in-memory hash table -- the lexicon. New
additions to the lexicon hash table are logged to a f ile. Once the words are converted into wordID's, their
occurrences in the current document are translated into hit lists and are written into the f orward barrels.
The main dif f iculty with parallelization of the indexing phase is that the lexicon needs to be shared.
Instead of sharing the lexicon, we took the approach of writing a log of all the extra words that were not
in a base lexicon, which we f ixed at 14 million words. That way multiple indexers can run in parallel and
then the small log f ile of extra words can be processed by one f inal indexer.
Sorting -- In order to generate the inverted index, the sorter takes each of the f orward barrels and
sorts it by wordID to produce an inverted barrel f or title and anchor hits and a f ull text inverted barrel.
This process happens one barrel at a time, thus requiring little temporary storage. Also, we parallelize
the sorting phase to use as many machines as we have simply by running multiple sorters, which can
process dif f erent buckets at the same time. Since the barrels don't f it into main memory, the sorter
f urther subdivides them into baskets which do f it into memory based on wordID and docID. Then the
sorter, loads each basket into memory, sorts it and writes its contents into the short inverted barrel and
the f ull inverted barrel.
4.5 Searching
The goal of searching is to provide quality search results ef f iciently. Many of the large commercial search
engines seemed to have made great progress in terms of ef f iciency. Theref ore, we have f ocused more on
quality of search in our research, although we believe our solutions are scalable to commercial volumes with a
bit more ef f ort. The google query evaluation process is show in Figure 4.

4. Scan through the doclists until there is a document that matches all the search terms.
5. Compute the rank of that document f or the query.
6. If we are in the short barrels and at the end of any doclist, seek to the start of the doclist in the f ull
barrel f or every word and go to step 4.
7. If we are not at the end of any doclist go to step 4.
Sort the documents that have matched by rank and return the top k.
Figure 4. Google Query Evaluation
To put a limit on response time, once a certain number (currently 40,000) of matching documents are f ound,
the searcher automatically goes to step 8 in Figure 4. This means that it is possible that sub-optimal results
would be returned. We are currently investigating other ways to solve this problem. In the past, we sorted the
hits according to PageRank, which seemed to improve the situation.
4.5.1 The Ranking System
Google maintains much more inf ormation about web documents than typical search engines. Every hitlist
includes position, f ont, and capitalization inf ormation. Additionally, we f actor in hits f rom anchor text and the
PageRank of the document. Combining all of this inf ormation into a rank is dif f icult. We designed our ranking
f unction so that no particular f actor can have too much inf luence. First, consider the simplest case -- a single
word query. In order to rank a document with a single word query, Google looks at that document's hit list f or
that word. Google considers each hit to be one of several dif f erent types (title, anchor, URL, plain text large
f ont, plain text small f ont, ...), each of which has its own type-weight. The type-weights make up a vector
indexed by type. Google counts the number of hits of each type in the hit list. Then every count is converted
into a count-weight. Count-weights increase linearly with counts at f irst but quickly taper of f so that more than
a certain count will not help. We take the dot product of the vector of count-weights with the vector of type-
weights to compute an IR score f or the document. Finally, the IR score is combined with PageRank to give a
f inal rank to the document.
For a multi-word search, the situation is more complicated. Now multiple hit lists must be scanned through at
once so that hits occurring close together in a document are weighted higher than hits occurring f ar apart. The
hits f rom the multiple hit lists are matched up so that nearby hits are matched together. For every matched set
of hits, a proximity is computed. The proximity is based on how f ar apart the hits are in the document (or
anchor) but is classif ied into 10 dif f erent value "bins" ranging f rom a phrase match to "not even close". Counts
are computed not only f or every type of hit but f or every type and proximity. Every type and proximity pair has a
type-prox-weight. The counts are converted into count-weights and we take the dot product of the count-
weights and the type-prox-weights to compute an IR score. All of these numbers and matrices can all be
displayed with the search results using a special debug mode. These displays have been very helpf ul in
developing the ranking system.
4.5.2 Feedback
The ranking f unction has many parameters like the type-weights and the type-prox-weights. Figuring out the
right values f or these parameters is something of a black art. In order to do this, we have a user f eedback
mechanism in the search engine. A trusted user may optionally evaluate all of the results that are returned. This
f eedback is saved. Then when we modif y the ranking f unction, we can see the impact of this change on all

Storage Statistics
Total Size of Fetched Pages 147.8 GB
Compressed Repository 53.5 GB
Short Inverted Index 4.1 GB
Full Inverted Index 37.2 GB
Lexicon 293 MB
Temporary Anchor Data
(not in total)
6.6 GB
Document Index Incl.
Variable Width Data
9.7 GB
previous searches which were ranked. Although f ar f rom perf ect, this gives us some idea of how a change in
the ranking f unction af f ects the search results.
5 Results and Performance
The most important measure of a search engine is the quality of its search results. While a complete user
evaluation is beyond the scope of this paper, our own experience with Google has shown it to produce better
results than the major commercial search engines f or most searches. As an example which illustrates the use
of PageRank, anchor text, and proximity, Figure 4 shows Google's results f or a search on "bill clinton". These
results demonstrates some of Google's f eatures. The results are clustered by server. This helps considerably
when sif ting through result sets. A number of results are f rom the whitehouse.gov domain which is what one
may reasonably expect f rom such a search. Currently, most major commercial search engines do not return any
results f rom whitehouse.gov, much less the right ones. Notice that there is no title f or the f irst result. This is
because it was not crawled. Instead, Google relied on anchor text to determine this was a good answer to the
query. Similarly, the f if th result is an email address which, of course, is not crawlable. It is also a result of
anchor text.
All of the results are reasonably high quality pages and, at last check, none were broken links. This is largely
because they all have high PageRank. The PageRanks are the percentages in red along with bar graphs. Finally,
there are no results about a Bill other than Clinton or about a Clinton other than Bill. This is because we place
heavy importance on the proximity of word occurrences. Of course a true test of the quality of a search engine
would involve an extensive user study or results analysis which we do not have room f or here. Instead, we
invite the reader to try Google f or themselves at http://google.stanf ord.edu.
5.1 Storage Requirements
Aside f rom search quality, Google is designed to scale cost ef f ectively to the size of the Web as it grows. One
aspect of this is to use storage ef f iciently. Table 1 has a breakdown of some statistics and storage
requirements of Google. Due to compression the total size of the repository is about 53 GB, just over one
third of the total data it stores. At current disk prices this makes the repository a relatively cheap source of
usef ul data. More importantly, the total of all the data used by the search engine requires a comparable amount
of storage, about 55 GB. Furthermore, most queries can be answered using just the short inverted index. With
better encoding and compression of the Document Index, a high quality web search engine may f it onto a 7GB
drive of a new PC.
5.2 System Performance
It is important f or a search engine to crawl and index
ef f iciently. This way inf ormation can be kept up to date and
major changes to the system can be tested relatively quickly.
For Google, the major operations are Crawling, Indexing, and
Sorting. It is dif f icult to measure how long crawling took
overall because disks f illed up, name servers crashed, or any
number of other problems which stopped the system. In total
it took roughly 9 days to download the 26 million pages
(including errors). However, once the system was running
smoothly, it ran much f aster, downloading the last 11 million
pages in just 63 hours, averaging just over 4 million pages
per day or 48.5 pages per second. We ran the indexer and the

Variable Width Data
Links Database 3.9 GB
Total Without Repository 55.2 GB
Total With Repository 108.7 GB
Web Page Statistics
Number of Web Pages Fetched 24 million
Number of Urls Seen 76.5 million
Number of Email Addresses 1.7 million
Number of 404's 1.6 million
Table 1. Statistics
Initial Query Same Query Repeated (IO mostly cached)
Query CPU Time(s) Total Time(s) CPU Time(s) Total Time(s)
al gore 0.09 2.13 0.06 0.06
vice president 1.77 3.84 1.66 1.80
hard disks 0.25 4.86 0.20 0.24
search engines 1.31 9.63 1.16 1.16
Table 2. Search Times
crawler simultaneously. The indexer ran just f aster than the
crawlers. This is largely because we spent just enough time
optimizing the indexer so that it would not be a bottleneck.
These optimizations included bulk updates to the document
index and placement of critical data structures on the local
disk. The indexer runs at roughly 54 pages per second. The
sorters can be run completely in parallel; using f our
machines, the whole process of sorting takes about 24
hours.
5.3 Search Performance
Improving the perf ormance of search was not the major
f ocus of our research up to this point. The current version of
Google answers most queries in between 1 and 10 seconds.
This time is mostly dominated by disk IO over NFS (since
disks are spread over a number of machines). Furthermore,
Google does not have any optimizations such as query
caching, subindices on common terms, and other common
optimizations. We intend to speed up Google considerably
through distribution and hardware, sof tware, and algorithmic improvements. Our target is to be able to handle
several hundred queries per second. Table 2 has some sample query times f rom the current version of Google.
They are repeated to show the speedups resulting f rom cached IO.
6
Conclusions
Google is designed to be a scalable search engine. The primary goal is to provide high quality search results
over a rapidly growing World Wide Web. Google employs a number of techniques to improve search quality
including page rank, anchor text, and proximity inf ormation. Furthermore, Google is a complete architecture f or
gathering web pages, indexing them, and perf orming search queries over them.
6.1 Future Work
A large-scale web search engine is a complex system and much remains to be done. Our immediate goals are to
improve search ef f iciency and to scale to approximately 100 million web pages. Some simple improvements to
ef f iciency include query caching, smart disk allocation, and subindices. Another area which requires much

research is updates. We must have smart algorithms to decide what old web pages should be recrawled and
what new ones should be crawled. Work toward this goal has been done in [Cho 98]. One promising area of
research is using proxy caches to build search databases, since they are demand driven. We are planning to
add simple f eatures supported by commercial search engines like boolean operators, negation, and stemming.
However, other f eatures are just starting to be explored such as relevance f eedback and clustering (Google
currently supports a simple hostname based clustering). We also plan to support user context (like the user's
location), and result summarization. We are also working to extend the use of link structure and link text.
Simple experiments indicate PageRank can be personalized by increasing the weight of a user's home page or
bookmarks. As f or link text, we are experimenting with using text surrounding links in addition to the link text
itself . A Web search engine is a very rich environment f or research ideas. We have f ar too many to list here so
we do not expect this Future Work section to become much shorter in the near f uture.
6.2 High Quality Search
The biggest problem f acing users of web search engines today is the quality of the results they get back.
While the results are of ten amusing and expand users' horizons, they are of ten f rustrating and consume
precious time. For example, the top result f or a search f or "Bill Clinton" on one of the most popular commercial
search engines was the Bill Clinton Joke of the Day: April 14, 1997. Google is designed to provide higher quality
search so as the Web continues to grow rapidly, inf ormation can be f ound easily. In order to accomplish this
Google makes heavy use of hypertextual inf ormation consisting of link structure and link (anchor) text. Google
also uses proximity and f ont inf ormation. While evaluation of a search engine is dif f icult, we have subjectively
f ound that Google returns higher quality search results than current commercial search engines. The analysis
of link structure via PageRank allows Google to evaluate the quality of web pages. The use of link text as a
description of what the link points to helps the search engine return relevant (and to some degree high quality)
results. Finally, the use of proximity inf ormation helps increase relevance a great deal f or many queries.
6.3 Scalable Architecture
Aside f rom the quality of search, Google is designed to scale. It must be ef f icient in both space and time, and
constant f actors are very important when dealing with the entire Web. In implementing Google, we have seen
bottlenecks in CPU, memory access, memory capacity, disk seeks, disk throughput, disk capacity, and network
IO. Google has evolved to overcome a number of these bottlenecks during various operations. Google's major
data structures make ef f icient use of available storage space. Furthermore, the crawling, indexing, and sorting
operations are ef f icient enough to be able to build an index of a substantial portion of the web -- 24 million
pages, in less than one week. We expect to be able to build an index of 100 million pages in less than a month.
6.4 A Research Tool
In addition to being a high quality search engine, Google is a research tool. The data Google has collected has
already resulted in many other papers submitted to conf erences and many more on the way. Recent research
such as [Abiteboul 97] has shown a number of limitations to queries about the Web that may be answered
without having the Web available locally. This means that Google (or a similar system) is not only a valuable
research tool but a necessary one f or a wide range of applications. We hope Google will be a resource f or
searchers and researchers all around the world and will spark the next generation of search engine technology.
7 Acknowledgments
Scott Hassan and Alan Steremberg have been critical to the development of Google. Their talented
contributions are irreplaceable, and the authors owe them much gratitude. We would also like to thank Hector
Garcia-Molina, Rajeev Motwani, Jef f Ullman, and Terry Winograd and the whole WebBase group f or their
support and insightf ul discussions. Finally we would like to recognize the generous support of our equipment

donors IBM, Intel, and Sun and our f unders. The research described here was conducted as part of the
Stanf ord Integrated Digital Library Project, supported by the National Science Foundation under Cooperative
Agreement IRI-9411306. Funding f or this cooperative agreement is also provided by DARPA and NASA, and by
Interval Research, and the industrial partners of the Stanf ord Digital Libraries Project.
References
[Abiteboul 97] Serge Abiteboul and Victor Vianu, Queries and Computation on the Web. Proceedings of
the International Conf erence on Database Theory. Delphi, Greece 1997.
[Bagdikian 97] Ben H. Bagdikian. The Media Monopoly. 5th Edition. Publisher: Beacon, ISBN: 0807061557
[Chakrabarti 98] S.Chakrabarti, B.Dom, D.Gibson, J.Kleinberg, P. Raghavan and S. Rajagopalan. Automatic
Resource Compilation by Analyzing Hyperlink Structure and Associated Text. Seventh International Web
Conf erence (WWW 98). Brisbane, Australia, April 14-18, 1998.
[Cho 98] Junghoo Cho, Hector Garcia-Molina, Lawrence Page. Efficient Crawling Through URL Ordering.
Seventh International Web Conf erence (WWW 98). Brisbane, Australia, April 14-18, 1998.
[Gravano 94] Luis Gravano, Hector Garcia-Molina, and A. Tomasic. The Effectiveness of GlOSS for the
Text-Database Discovery Problem. Proc. of the 1994 ACM SIGMOD International Conf erence On
Management Of Data, 1994.
[Kleinberg 98] Jon Kleinberg, Authoritative Sources in a Hyperlinked Environment, Proc. ACM-SIAM
Symposium on Discrete Algorithms, 1998.
[Marchiori 97] Massimo Marchiori. The Quest for Correct Information on the Web: Hyper Search Engines.
The Sixth International WWW Conf erence (WWW 97). Santa Clara, USA, April 7-11, 1997.
[McBryan 94] Oliver A. McBryan. GENVL and WWWW: Tools for Taming the Web. First International
Conference on the World Wide Web. CERN, Geneva (Switzerland), May 25-26-27 1994.
http://www.cs.colorado.edu/home/mcbryan/mypapers/www94.ps
[Page 98] Lawrence Page, Sergey Brin, Rajeev Motwani, Terry Winograd. The PageRank Citation Ranking:
Bringing Order to the Web. Manuscript in progress. http://google.stanf ord.edu/~backrub/pageranksub.ps
[Pinkerton 94] Brian Pinkerton, Finding What People Want: Experiences with the WebCrawler. The Second
International WWW Conf erence Chicago, USA, October 17-20, 1994.
http://inf o.webcrawler.com/bp/WWW94.html
[Spertus 97] Ellen Spertus. ParaSite: Mining Structural Information on the Web. The Sixth International
WWW Conf erence (WWW 97). Santa Clara, USA, April 7-11, 1997.
[TREC 96] Proceedings of the fifth Text REtrieval Conference (TREC-5). Gaithersburg, Maryland,
November 20-22, 1996. Publisher: Department of Commerce, National Institute of Standards and
Technology. Editors: D. K. Harman and E. M. Voorhees. Full text at: http://trec.nist.gov/
[Witten 94] Ian H Witten, Alistair Mof f at, and Timothy C. Bell. Managing Gigabytes: Compressing and
Indexing Documents and Images. New York: Van Nostrand Reinhold, 1994.
[Weiss 96] Ron Weiss, Bienvenido Velez, Mark A. Sheldon, Chanathip Manprempre, Peter Szilagyi,
Andrzej Duda, and David K. Gif f ord. HyPursuit: A Hierarchical Network Search Engine that Exploits
Content-Link Hypertext Clustering. Proceedings of the 7th ACM Conf erence on Hypertext. New York,
1996.
Vitae

Sergey Brin received his B.S. degree in mathematics and computer science f rom the
University of Maryland at College Park in 1993. Currently, he is a Ph.D. candidate in
computer science at Stanf ord University where he received his M.S. in 1995. He is a
recipient of a National Science Foundation Graduate Fellowship. His research
interests include search engines, inf ormation extraction f rom unstructured sources,
and data mining of large text collections and scientif ic data.
Lawrence Page was born in East Lansing, Michigan, and received a B.S.E. in
Computer Engineering at the University of Michigan Ann Arbor in 1995. He is currently
a Ph.D. candidate in Computer Science at Stanf ord University. Some of his research
interests include the link structure of the web, human computer interaction, search
engines, scalability of inf ormation access interf aces, and personal data mining.
8 Appendix A: Advertising and Mixed Motives
Currently, the predominant business model f or commercial search engines is
advertising. The goals of the advertising business model do not always
correspond to providing quality search to users. For example, in our prototype
search engine one of the top results f or cellular phone is "The Ef f ect of Cellular
Phone Use Upon Driver Attention", a study which explains in great detail the
distractions and risk associated with conversing on a cell phone while driving. This
search result came up f irst because of its high importance as judged by the
PageRank algorithm, an approximation of citation importance on the web [Page,
98]. It is clear that a search engine which was taking money f or showing cellular
phone ads would have dif f iculty justif ying the page that our system returned to its
paying advertisers. For this type of reason and historical experience with other
media [Bagdikian 83], we expect that advertising f unded search engines will be inherently biased towards the
advertisers and away f rom the needs of the consumers.
Since it is very dif f icult even f or experts to evaluate search engines, search engine bias is particularly insidious.
A good example was OpenText, which was reported to be selling companies the right to be listed at the top of
the search results f or particular queries [Marchiori 97]. This type of bias is much more insidious than
advertising, because it is not clear who "deserves" to be there, and who is willing to pay money to be listed.
This business model resulted in an uproar, and OpenText has ceased to be a viable search engine. But less
blatant bias are likely to be tolerated by the market. For example, a search engine could add a small f actor to
search results f rom "f riendly" companies, and subtract a f actor f rom results f rom competitors. This type of
bias is very dif f icult to detect but could still have a signif icant ef f ect on the market. Furthermore, advertising
income of ten provides an incentive to provide poor quality search results. For example, we noticed a major
search engine would not return a large airline's homepage when the airline's name was given as a query. It so
happened that the airline had placed an expensive ad, linked to the query that was its name. A better search
engine would not have required this ad, and possibly resulted in the loss of the revenue f rom the airline to the
search engine. In general, it could be argued f rom the consumer point of view that the better the search engine
is, the f ewer advertisements will be needed f or the consumer to f ind what they want. This of course erodes
the advertising supported business model of the existing search engines. However, there will always be money
f rom advertisers who want a customer to switch products, or have something that is genuinely new. But we
believe the issue of advertising causes enough mixed incentives that it is crucial to have a competitive search
engine that is transparent and in the academic realm.
9 Appendix B: Scalability
9. 1 Scalability of Google

9. 1 Scalability of Google
We have designed Google to be scalable in the near term to a goal of 100 million web pages. We have just
received disk and machines to handle roughly that amount. All of the time consuming parts of the system are
parallelize and roughly linear time. These include things like the crawlers, indexers, and sorters. We also think
that most of the data structures will deal gracef ully with the expansion. However, at 100 million web pages we
will be very close up against all sorts of operating system limits in the common operating systems (currently we
run on both Solaris and Linux). These include things like addressable memory, number of open f ile descriptors,
network sockets and bandwidth, and many others. We believe expanding to a lot more than 100 million pages
would greatly increase the complexity of our system.
9.2 Scalability of Centralized Indexing Architectures
As the capabilities of computers increase, it becomes possible to index a very large amount of text f or a
reasonable cost. Of course, other more bandwidth intensive media such as video is likely to become more
pervasive. But, because the cost of production of text is low compared to media like video, text is likely to
remain very pervasive. Also, it is likely that soon we will have speech recognition that does a reasonable job
converting speech into text, expanding the amount of text available. All of this provides amazing possibilities
f or centralized indexing. Here is an illustrative example. We assume we want to index everything everyone in the
US has written f or a year. We assume that there are 250 million people in the US and they write an average of
10k per day. That works out to be about 850 terabytes. Also assume that indexing a terabyte can be done now
f or a reasonable cost. We also assume that the indexing methods used over the text are linear, or nearly linear
in their complexity. Given all these assumptions we can compute how long it would take bef ore we could index
our 850 terabytes f or a reasonable cost assuming certain growth f actors. Moore's Law was def ined in 1965 as
a doubling every 18 months in processor power. It has held remarkably true, not just f or processors, but f or
other important system parameters such as disk as well. If we assume that Moore's law holds f or the f uture,
we need only 10 more doublings, or 15 years to reach our goal of indexing everything everyone in the US has
written f or a year f or a price that a small company could af f ord. Of course, hardware experts are somewhat
concerned Moore's Law may not continue to hold f or the next 15 years, but there are certainly a lot of
interesting centralized applications even if we only get part of the way to our hypothetical example.
Of course a distributed systems like Gloss [Gravano 94] or Harvest will of ten be the most ef f icient and elegant
technical solution f or indexing, but it seems dif f icult to convince the world to use these systems because of
the high administration costs of setting up large numbers of installations. Of course, it is quite likely that
reducing the administration cost drastically is possible. If that happens, and everyone starts running a
distributed indexing system, searching would certainly improve drastically.
Because humans can only type or speak a f inite amount, and as computers continue improving, text indexing
will scale even better than it does now. Of course there could be an inf inite amount of machine generated
content, but just indexing huge amounts of human generated content seems tremendously usef ul. So we are
optimistic that our centralized web search engine architecture will improve in its ability to cover the pertinent
text inf ormation over time and that there is a bright f uture f or search.

History page-brin thesis - anatomy of a large scale hypertextual web search engine

More Related Content

What's hot (9)

Viewers also liked (14)

Similar to History page-brin thesis - anatomy of a large scale hypertextual web search engine (20)

More from Bitsytask (20)

Recently uploaded (20)

History page-brin thesis - anatomy of a large scale hypertextual web search engine