Ecient Crawling Through URL Ordering

Junghoo Cho Hector Garcia-Molina Lawrence Page

Department of Computer Science
Stanford, CA 94305
{cho, hector, page}@cs.stanford.edu
Abstract
In this paper we study in what order a crawler should visit the URLs it has seen, in order
to obtain more important pages rst. Obtaining important pages rapidly can be very useful
when a crawler cannot visit the entire Web in a reasonable amount of time. We dene several
importance metrics, ordering schemes, and performance evaluation measures for this problem.
We also experimentally evaluate the ordering schemes on the Stanford University Web. Our
results show that a crawler with a good ordering scheme can obtain important pages signicantly
faster than one without.
1 Introduction
A crawler is a program that retrieves Web pages, commonly for use by a search engine [Pin94] or
a Web cache. Roughly, a crawler starts o with the URL for an initial page p
0
. It retrieves p
0
,
extracts any URLs in it, and adds them to a queue of URLs to be scanned. Then the crawler
gets URLs from the queue (in some order), and repeats the process. Every page that is scanned
is given to a client that saves the pages, creates an index for the pages, or summarizes or analyzes
the content of the pages.
Crawlers are widely used today. Crawlers for the major search engines (e.g., AltaVista, InfoSeek,
Excite, and Lycos) attempt to visit most text Web pages, in order to build content indexes. Other
crawlers may also visit many pages, but may look only for certain types of information (e.g., email
addresses). At the other end of the spectrum, we have personal crawlers that scan for pages of
interest to a particular user, in order to build a fast access cache (e.g. NetAttche).
The design of a good crawler presents many challenges. Externally, the crawler must avoid
overloading Web sites or network links as it goes about its business [Kos95]. Internally, the crawler
must deal with huge volumes of data. Unless it has unlimited computing resources and unlimited
time, it must carefully decide what URLs to scan and in what order. The crawler must also decide
how frequently to revisit pages it has already seen, in order to keep its client informed of changes
on the Web. Inspite of all these challenges, and the importance of crawlers on the Internet, very
little research has been done on crawlers.
In this paper we address one of these important challenges: How should a crawler select URLs
to scan from its queue of known URLs? If a crawler intends to perform a single scan of the entire
Web, and the load placed on target sites is not an issue, then any URL order will suce. That
is, eventually every single known URL will be visited, so the order is not critical. However, most
crawlers will not be able to visit every possible page for two main reasons:
1
Their client may have limited storage capacity, and may be unable to index or analyze all
pages. Currently the Web contains about 1.5TB and is growing rapidly, so it is reasonable to
expect that most clients will not want or will not be able to cope with all that data [Kah97].
Crawling takes time, so at some point the crawler may need to start revisiting previously
scanned pages, to check for changes. This means that it may never get to some pages. It is
currently estimated that over 600GB of the Web changes every month [Kah97].
In either case, it is important for the crawler to visit important pages rst, so that the fraction
of the Web that is visited (and kept up to date) is more meaningful. In the following sections, we
present several dierent useful denitions of importance, and develop crawling priorities so that
important pages have a higher probability of being visited rst. We also present experimental
results from crawling the Stanford University Web pages that show how eective the dierent
crawling strategies are.
2 Importance metrics
Not all pages are necessarily of equal interest to a crawlers client. For instance, if the client is
building a specialized database on a particular topic, then pages that refer to that topic are more
important, and should be visited as early as possible. Similarly, a search engine may use the number
of Web URLs that point to a page, the so-called backlink count, to rank user query results. If the
crawler cannot visit all pages, then it is better to visit those with a high backlink count, since this
will give the end-user higher ranking results.
Given a Web page p, we can dene the importance of the page, I(p), in one of the following
ways. (These metrics can be combined, as will be discussed later.)
1. Similarity to a Driving Query Q: A query Q drives the crawling process, and I(p) is dened to
be the textual similarity between p and Q. Similarity has been well studied in the Information
Retrieval (IR) community [Sal83] and has been applied to the Web environment [YLYL95].
We use IS(p) to refer to the importance metric in this case. We also use IS(p, Q) when we
wish to make the query explicit.
To compute similarities, we can view each document (p or Q) as an n-dimensional vector
w
1
, . . . , w
n
. The term w
i
in this vector represents the ith word in the vocabulary. If w
i
does not appear in the document, then w
i
is zero. If it does appear, w
i
is set to represent the
signicance of the word. One common way to compute the signicance w
i
is to multiply the
number of times the ith word appears in the document by the inverse document frequency
(idf) of the ith word. The idf factor is one divided by the number of times the word appears in
the entire collection, which in this case would be the entire Web. The idf factor corresponds
to the content discriminating power of a word: A term that appears rarely in documents (e.g.,
queue) has a high idf, while a term that occurs in many documents (e.g., the) has a low
idf. (The w
i
terms can also take into account where in a page the word appears. For instance,
words appearing in the title of an HTML page may be given a higher weight than other words
in the body.) The similarity between p and Q can then be dened as the inner product of the
p and Q vectors. Another option is to use the cosine similarity measure, which is the inner
product of the normalized vectors.
2
Note that if we do not use idf terms in our similarity computation, the importance of a page,
IS(p), can be computed with local information, i.e., just p and Q. However, if we use idf
terms, then we need global information. During the crawling process we have not seen the
entire collection, so we have to estimate the idf factors from the pages that have been crawled,
or from some reference idf terms computed at some other time. We use IS

(p) to refer to
the estimated importance of page p, which is dierent from the actual importance IS(p) that
is computable only after the entire Web has been crawled. If idf factors are not used, then
IS

(p) = IS(p).
2. Backlink Count: The value of I(p) is the number of links to p that appear over the entire
Web. We use IB(p) to refer to this importance metric. Intuitively, a page p that is linked to
by many pages is more important than one that is seldom referenced. This type of citation
count has been used extensively to evaluate the impact of published papers. On the Web,
IB(p) is useful for ranking query results, giving end-users pages that are more likely to be of
general interest.
Note that evaluating IB(p) requires counting backlinks over the entire Web. A crawler may
estimate this value with IB

(p), the number of links to p that have been seen so far.

3. PageRank: The IB(p) metric treats all links equally. Thus, a link from the Yahoo home page
counts the same as a link from some individuals home page. However, since the Yahoo home
page is more important (it has a much higher IB count), it would make sense to value that
link more highly. The PageRank backlink metric, IR(p), recursively denes the importance
of a page to be the weighted sum of the importance of the pages that have backlinks to p.
Such a metric has been found to be very useful in ranking results of user queries [PB98, Goo].
We use IR

(p) for the estimated value of IR(p) when we have only a subset of pages available.
More formally, if a page has no outgoing link, we assume that it has outgoing links to every
single Web page. Next, consider a page p that is pointed at by pages t
1
, . . . , t
n
. Let c
i
be
the number of links going out of page t
i
. Also, let d be a damping factor (whose intuition is
given below). Then, the weighted backlink count of page p is given by
IR(p) = (1 d) +d [IR(t
1
)/c
1
+ +IR(t
n
)/c
n
]
This leads to one equation per Web page, with an equal number of unknowns. The equations
can be solved for the IR values. They can be solved iteratively, starting with all IR values
equal to 1. At each step, the new IR(p) value is computed from the old IR(t
i
) values (using
the equation above), until the values converge. This calculation corresponds to computing
the principal eigenvector of the link matrices.
One intuitive model for PageRank is that we can think of a user surng the Web, starting
from any page, and randomly selecting from that page a link to follow. When the user reaches
a page with no outlinks, he jumps to a random page. Also, when the user is on a page, there
is some probability, d, that the next visited page will be completely random. This damping
factor d makes sense because users will only continue clicking on one task for a nite amount
of time before they go on to something unrelated. The IR(p) values we computed above give
us the probability that our random surfer is at p at any given time.
3
4. Forward Link Count: For completeness we may want to consider a metric IF(p) that counts
the number of links that emanate from p. Under this metric, a page with many outgoing links
is very valuable, since it may be a Web directory. This metric can be computed directly from
p, so IF

(p) = IF(p). This kind of metric has been used in conjunction with other factors
to reasonably identify index pages [PPR96]. We could also dene a weighted forward link
metric, analogous to IR(p), but we do not consider it here.
5. Location Metric: The IL(p) importance of page p is a function of its location, not of its
contents. If URL u leads to p, then IL(p) is a function of u. For example, URLs ending with
.com may be deemed more useful than URLs with other endings, or URLs containing the
string home may be more of interest than other URLs. Another location metric that is
sometimes used considers URLs with fewer slashes more useful than those with more slashes.
All these examples are local metrics since they can be evaluated simply by looking at the
URL u.
As stated earlier, our importance metrics can be combined in various ways. For example, we
may dene a metric IC(p) = k
1
IS(p, Q) +k
2
IB(p), for some constants k
1
, k
2
. This combines the
similarity metric (under some given query Q) and the backlink metric. Pages that have relevant
content and many backlinks would be the highest ranked. (Note that a similar approach was used
to improve the eectiveness of a search engine [Mar97].)
3 Problem denition
Our goal is to design a crawler that if possible visits high I(p) pages before lower ranked ones, for
some denition of I(p). Of course, the crawler will only have available I

(p) values, so based on

these it will have to guess what are the high I(p) pages to fetch next.
Our general goal can be stated more precisely in three ways, depending on how we expect the
crawler to operate. (In our evaluations of Section 5 we use the second model in most cases, but
we do compare it against the rst model in one experiment. Nevertheless, we believe it is useful to
discuss all three models to understand the options.)
Crawl & Stop: Under this model, the crawler C starts at its initial page p
0
and stops after
visiting K pages. At this point a perfect crawler would have visited pages r
1
, . . . , r
K
, where r
1
is the page with the highest importance value, r
2
is the next highest, and so on. We call pages
r
1
through r
K
the hot pages. The K pages visited by our real crawler will contain only
M pages with rank higher than or equal to I(r
K
). We dene the performance of the crawler
C to be P
CS
(C) = M/K. The performance of the ideal crawler is of course 1. A crawler
that somehow manages to visit pages entirely at random, and may revisit pages, would have
a performance of K/T, where T is the total number of pages in the Web. (Each page visited
is a hot page with probability K/T. Thus, the expected number of desired pages when the
crawler stops is K
2
/T.)
Crawl & Stop with Threshold: We again assume that the crawler visits K pages. How-
ever, we are now given an importance target G, and any page with I(p) G is considered
hot. Let us assume that the total number of hot pages is H. The performance of the crawler,
P
ST
(C), is the fraction of the H hot pages that have been visited when the crawler stops. If
4
K < H , then an ideal crawler will have performance K/H. If K H, then the ideal crawler
has the perfect performance 1. A purely random crawler that revisits pages is expected to
visit (H/T) K hot pages when it stops. Thus, its performance is K/T. Only when the
random crawler visits all T pages its performance is expected to be 1.
Limited Buer Crawl: In this model we consider the impact of limited storage on the
crawling process. We assume that the crawler can only keep B pages in its buer. Thus,
after the buer lls up, the crawler must decide what pages to ush to make room for new
pages. An ideal crawler could simply drop the pages with lowest I(p) value, but a real crawler
must guess which of the pages in its buer will eventually have low I(p) values. We allow the
crawler to visit a total of T pages, equal to the total number of Web pages. At the end of this
process, the fraction of the B buer pages that are hot gives us the performance P
BC
(C). We
can dene hot pages to be those with I(p) G, where G is a target importance, or those with
I(p) I(r
B
), where r
B
is the page with the Bth highest importance value. The performance
of an ideal and a random crawler are analogous to those in the previous cases.
Note that to evaluate a crawler under any of these metrics, we need to compute the actual I(p)
values of pages, and this involves crawling the entire Web. To keep our experiments (Section 5)
manageable, we imagines that the Stanford University pages form the entire Web, and we only
evaluate performance in this context. That is, we assume that all pages outside of Stanford have
I(p) = 0, and that links to pages outside of Stanford or links from pages outside of Stanford
do not count in I(p) computations. In Section 5.2 we study the implications of this assumption
by also analyzing a smaller Web within the Stanford domain, and seeing how Web size impacts
performance.
4 Ordering metrics
A crawler keeps a queue of URLs it has seen during a crawl, and must select from this queue the
next URL to visit. The ordering metric O is used by the crawler for this selection, i.e., it selects
the URL u such that O(u) has the highest value among all URLs in the queue. The O metric can
only use information seen (and remembered if space is limited) by the crawler.
The O metric should be designed with an importance metric in mind. For instance, if we are
searching for high IB(p) pages, it makes sense to use O(u) = IB

(p), where p is the page u points

to. However, it might also make sense to use O(u) = IR

(p), even if our importance metric is not

weighted. In our experiments, we explore the types of ordering metrics that are best suited for
either IB(p) or IR(p).
For a location importance metric IL(p), we can use that metric directly for ordering since the
URL of p directly gives the IL(p) value. However, for forward link IF(p) and similarity IS(p)
metrics, it is much harder to devise an ordering metric since we have not seen p yet. As we will see,
for similarity, we may be able to use the text that anchors the URL u as a predictor of the text
that p might contain. Thus, one possible ordering metric O(u) is IS(A, Q), where A is the anchor
text of the URL u, and Q is the driving query.
5
5 Experiments
To avoid network congestion and heavy loads on the servers, we did our experimental evaluation
in two steps. In the rst step, we physically crawled all Stanford Web pages and built a local
repository of the pages. This was done with the Stanford WebBase [BP98], a system designed to
create and maintain large Web repositories.
After we built the repository, we ran our virtual crawlers on it to evaluate dierent crawling
schemes. Note that even though we had the complete image of the Stanford domain in the repos-
itory, our virtual crawler based its crawling decisions only on the pages it saw for itself. In this
section we briey discuss how the particular database was obtained for our experiments.
5.1 Description of dataset
To download an image of the Stanford Web pages, we started WebBase with an initial list of
stanford.edu URLs. These 89,119 URLs were obtained from an earlier crawl. During the
crawl, non-Stanford URLs were ignored. Also, we limited the actual data that we collected for
two reasons. The rst is that many heuristics are needed to avoid automatically generated, and
potentially innite, sets of pages. For example, any URLs containing /cgi-bin/ are not crawled,
because they are likely to contain programs which generate innite sets of pages, or produce other
undesirable side eects such as an unintended vote in an online election. Several other heuristics
similar to aspects discussed in the location metric described earlier are used to eliminate URLs which
look undesirable. Another way the data set is reduced is through the robots exclusion protocol [Rob],
which allows Webmasters to dene pages they do not want crawled by automatic systems.
At the end of the process, we had 784,592 known URLs to Stanford pages. It should be noted
that 352,944 of the known URLs were on one server, http://www.slac.stanford.edu, which has
a program that could generate an unlimited number of Web pages. The crawl was stopped before
it was complete, but most of the uncrawled URLs were on only a few servers, so we believe the
dataset we used to be a reasonable representation of the stanford.edu Web. This dataset consisted
of about 225,000 crawled valid HTML pages,
1
which consumed roughly 2.5GB of disk space.
However, out of these 225,000 pages, 46,000 pages were unreachable from the starting point of the
crawl, so the total number of pages for our experiments was 179,000.
We should stress that the virtual crawlers that will be discussed next do not use WebBase
directly. As stated earlier, they use the dataset collected by the WebBase crawler, and do their
own crawling on it. The virtual crawlers are simpler than the WebBase crawler. For instance, they
can detect if a URL is invalid simply by seeing if it is in the dataset. Similarly, they do not need
to distribute the load to visited sites. These simplications are ne, since the virtual crawlers are
only used to evaluate ordering schemes, and not to do real crawling.
5.2 Backlink-based crawlers
In this section we study the eectiveness of various ordering metrics, for the scenario where im-
portance is measured through backlinks (i.e., either the IB(p) or IR(p) metrics). We start by
describing the structure of the virtual crawler, and then consider the dierent ordering metrics.
Unless otherwise noted, we use the Stanford dataset described in Section 5.1, and all crawls are
1
We considered a page valid when its Web server responded with the HTTP header 200 OK.
6
Algorithm 5.1 Crawling algorithm (backlink based)
Input: starting url: seed URL
Procedure:
[1] enqueue(url queue, starting url)
[2] while (not empty(url queue))
[3] url = dequeue(url queue)
[4] page = crawl page(url)
[5] enqueue(crawled pages, (url, page))
[6] url list = extract urls(page)
[7] foreach u in url list
[8] enqueue(links, (url, u))
[9] if (u/ url queue and (u,-)/ crawled pages)
[10] enqueue(url queue, u)
[11] reorder queue(url queue)
Function description:
enqueue(queue, element): append element at the end of queue
dequeue(queue) : remove the element at the beginning
of queue and return it
reorder queue(queue) : reorder queue using information in
links (refer to Figure 2)
Figure 1: Basic crawling algorithm
started from the Stanford homepage. For the PageRank metric we use a damping factor d of 0.9
(for both IR(p) and IR

(p)) for all of our experiments.

Figure 1 shows our basic virtual crawler. The crawler manages three main data structures.
Queue url queue contains the URLs that have been seen and need to be visited. Once a page
is visited, it is stored (with its URL) in crawled pages. Finally, links holds pairs of the form
(u
1
, u
2
), where URL u
2
was seen in the visited page with URL u
1
. The crawlers ordering metric
is implemented by the function reorder queue(), shown in Figure 2. We used three ordering
metrics: (1) breadth-rst (2) backlink count IB

(p), and (3) PageRank IR

(p). The breadth-rst

metric places URLs in the queue in the order in which they are discovered, and this policy makes
the crawler visit pages in breadth-rst order.
We start by showing in Figure 3 the crawlers performance with the backlink ordering metric.
In this scenario, the importance metric is the number of backlinks to a page (I(p) = IB(p)) and
we consider a Crawl & Stop with Threshold model in Section 3 with G either 3, 10, or 100. Recall
that a page with G or more backlinks is considered important, i.e., hot. Under these hot page
denitions, about H = 85,000 (47%), 17,500 (10%) and 1,400 (0.8%) pages out of 179,000 total
Web pages were considered hot, respectively.
In Figure 3, the horizontal axis is the fraction of the Stanford Web pages that has been crawled
over time. At the right end of the horizontal axis, all 179,000 pages have been visited. The vertical
axis represents P
ST
, the fraction of the total hot pages that has been crawled at a given point. The
7
(1) breadth rst
do nothing (null operation)
(2) backlink count, IB

(p)
foreach u in url queue
backlink count[u] = number of terms (-,u) in links
sort url queue by backlink count[u]
(3) PageRank IR

(p)
solve the following set of equations:
IR[u] = (1 0.9) + 0.9

i
IR[v
i
]
c
i
, where
(v
i
, u) links and c
i
is the number of links in the page v
i
sort url queue by IR(u)
Figure 2: Description of reorder queue() of each ordering metric
0.2 0.4 0.6 0.8 1
Fraction of Stanford Web crawled
0.2
0.4
0.6
0.8
1
P
experiment ideal
G=10
G=3
G=100
S
T
Figure 3: Fraction of Stanford Web crawled vs. P
ST
. I(p) = IB(p); O(u) = IB

(p).
8
0.2 0.4 0.6 0.8 1
Fraction of Stanford Web crawled
0.2
0.4
0.6
0.8
1
P
S
T
Ordering metric O(u)
Breadth
PageRank
Random
Ideal
Backlink
Figure 4: Fraction of Stanford Web crawled vs. P
ST
. I(p) = IB(p); G = 100.
solid lines in the gure show the results from our experiments. For example, when the crawler in
our experiment visited 0.2 (20%) of the Stanford pages, it crawled 0.5 (50%) of the total hot pages
for G = 100. The dashed lines in the graph show the expected performance of ideal crawlers. An
ideal crawler reaches performance 1 when H pages have been crawled. The dotted line represents
the performance of a random crawler, and it increases linearly over time.
The graph shows that as our denition of a hot page becomes more stringent (larger G), the
faster the crawler can locate the hot pages. This result is to be expected, since pages with many
backlinks will be seen quickly after the crawl starts. Figure 3 also shows that even if G is large,
nding the last group of hot pages is always dicult. That is, to the right of the 0.8 point on
the horizontal axis, the crawler nds hot pages at roughly the same rate as a random crawler.
In our next experiment we compare three dierent ordering metrics: 1) breadth-rst 2) backlink-
count and 3) PageRank (corresponding to the three functions of Figure 2). We continue to use the
Crawl & Stop with Threshold model, with G = 100, and a IB(p) importance metric. Figure 4 shows
the results of this experiment. The results are rather counterintuitive. Intuitively one would expect
that a crawler using the backlink ordering metric IB

(p) that matches the importance metric IB(p)

would perform the best. However, this is not the case, and the PageRank metric IR

(p) outperforms
the IB

(p) one. To understand why, we manually traced the crawlers operation. We noticed that
often the IB

(p) crawler behaved like a depth-rst one, frequently visiting pages in one cluster
before moving on to the next. On the other hand, the IR

(p) crawler combined breadth and depth

in a better way.
To illustrate, let us consider the Web fragment of Figure 5. With IB

(p) ordering, the crawler

visits a page like the one labeled p
1
and quickly nds a cluster A of pages that point to each other.
The A pages temporarily have more backlinks than page p
2
, so the visit of page p
2
is delayed even
if page p
2
actually has more backlinks than the pages in cluster A. On the other hand, with IR

(p)
ordering, page p
2
may have higher rank (because its link comes from a high ranking page) than the
pages in cluster A (that only have pointers from low ranking pages within the cluster). Therefore,
page p
2
is reached faster.
In summary, during the early stages of a crawl, the backlink information is biased by the starting
point. If the crawler bases its decisions on this skewed information, it tries getting locally hot pages
instead of globally hot pages, and this bias gets worse as the crawl proceeds. On the other hand,
9
p
1
p
2
Cluster B Cluster A
Figure 5: Crawling order
0.2 0.4 0.6 0.8 1
Fraction of Stanford Web crawled
0.2
0.4
0.6
0.8
1
P
Ordering metric O(u)
Breadth
PageRank
Random
Ideal
Backlink C
S
Figure 6: Fraction of Stanford Web crawled vs. P
CS
. I(p) = IB(p).
the IR

(p) PageRank crawler is not as biased towards locally hot pages, so it gives better results
regardless of the starting point.
Figure 6 shows that this conclusion is not limited to the Crawl & Stop with Threshold model.
In the gure we show the performance of the crawlers under the Crawl & Stop model (Section 3).
Remember that under the Crawl & Stop model, the denition of hot pages changes over time. That
is, the crawler does not have a predened notion of hot pages, and instead, when the crawler has
visited, say, 30% of the entire Web, it considers the top 30% pages as hot pages. Therefore, an
ideal crawler would have performance 1 at all times because it would download pages in the order
of their importance. Figure 6 compares 1) breadth-rst 2) backlink and 3) PageRank ordering
metrics for the IB(p) importance metric under this model. The vertical axis represents P
CS
, the
crawled fraction of hot pages at each point under the varying denition of hot pages. The gure
shows that the results of the Crawl & Stop model are analogous to those of the Crawl & Stop with
Threshold model: The PageRank ordering metric shows the best performance.
Returning to the Crawl & Stop with Threshold model, Figure 7 shows the results when we use
the IR(p) PageRank importance metric with G = 13.
2
Again, the PageRank ordering metric shows
2
When G = 13 for the IR(p) metric, the number of hot pages was about 1, 700 (1%), which is close to the 1, 400
10
0.2 0.4 0.6 0.8 1
Fraction of Stanford Web crawled
0.2
0.4
0.6
0.8
1
P
Ordering metric O(u)
Breadth
PageRank
Random
Ideal
Backlink S
T
Figure 7: Fraction of Stanford Web crawled vs. P
ST
. I(p) = IR(p); G = 13.
0.2 0.4 0.6 0.8 1
Fraction of Database Group Web crawled
0.2
0.4
0.6
0.8
1
P
S
T
Ordering metric O(u)
Breadth
PageRank
Random
Ideal
Backlink
Figure 8: Fraction of DB group Web crawled vs. P
ST
. I(p) = IB(p); G = 5.
the best performance. The backlink and the breadth-rst metrics show similar performance. Based
on these results, we recommend using the PageRank ordering metric for both the IB(p) and the
IR(p) importance metrics.
5.3 Small-scale crawl
In some cases the crawlers client may only be interested in small portions of the Web. For instance,
the client may be interested in a single site (to create a mirror, say). In this subsection we evaluate
the ordering metrics in such a scenario.
To study the impact of scale on the performance of a crawler we ran experiments similar to
those in Section 5.2 only on the pages of the Stanford Database Group (on the server http:
//www-db.stanford.edu). This subset of the Stanford pages consists of about 1,100 HTML pages,
which is much smaller than the entire Stanford domain. In most of our experiments for the Database
Group domain, crawling performance was not as good as for the Stanford domain. Figure 8 shows
of Figure 4.
11
1 2 3 4 5 6 7 8 9 10111213141516171819
100
200
300
400
500
N
u
m
b
e
r

o
f

p
a
g
e
s
Number of backlinks
Figure 9: Histogram of backlink counts within DB group Web
one of the results. In this case, we use the Crawl & Stop with Threshold model with G = 5 with
the importance metric IB(p). The graph shows that performance can be even worse than that of
a random crawler at times, for all ordering metrics.
This poor performance is mainly because an importance metric based on backlinks is not a
good measure of importance for a small domain. In a small domain, most pages have only a small
number of backlinks and the number of backlinks therefore is very sensitive to a page creators
style. For example, Figure 9 shows the histogram for the number of backlinks in the Database
Group domain. The vertical axis shows the number of pages with a given backlink count. We can
see that most pages have fewer than 5 backlinks. In this range, the rank of each page varies greatly
according to the style used by the creator of the page. If the creator generates many cross-links
between his pages, then his pages have a high IB(p) rank, otherwise they do not. Therefore, the
rank is very sensitive and is not a good measure of the importance of the pages.
In Figure 8 we can see the impact of locally dense clusters that have many locally popular
but globally unpopular pages. The performance of the backlink IB

(p) crawler is initially quite

at: It initially does a depth-rst crawl for the rst cluster it found. After visiting about 20% of the
pages, the crawler suddenly discovers a large cluster, and this accounts for the jump in the graph.
On the other hand, the PageRank IR

(p) crawler found this large cluster earlier, so its performance

is much better initially.
In Figure 10 we show the results on the Database Group Web when the importance metric is
IR(p) PageRank metric with G = 3.
3
All three ordering metrics show better performance under
the IR(p) metric than under the IB(p) metric, but still performance is not as good as that of the
larger Stanford domain. Again, the IR

(p) ordering metric shows the best performance.

5.4 Similarity-based crawlers
In the experiments of two previous subsections, we compared three dierent backlink-based crawlers.
In this subsection, we present the results of our experiments on similarity-based crawlers. The
similarity-based importance metric, IS(p), measures the relevance of each page to a topic or a
query that the user has in mind. There are clearly many possible IS(p) metrics to consider, so
our experiments here are not intended to be comprehensive. Instead, our goal is to briey explore
3
When G = 3, the number of hot pages is similar to that of Figure 8.
12
0.2 0.4 0.6 0.8 1
Fraction of Database Group Web crawled
0.2
0.4
0.6
0.8
1
P
S
T
Ordering metric O(u)
Breadth
PageRank
Random
Ideal
Backlink
Figure 10: Percentage of DB group Web crawled vs. P
ST
. I(p) = IR(p); G = 3.
the potential of various ordering schemes in some sample scenarios. In particular, for our rst two
experiments we consider the following IS(p) denition: A page is considered hot if it contains the
word computer in its title or if it has more than 10 occurrences of computer in its body.
4
For similarity-based crawling, the crawler of Figure 1 is not appropriate, since it does not
take the content of the page into account. To give priority to the pages related to the topic of
interest, we modied our crawler as shown in Figure 11. This crawler keeps two queues of URLs
to visit: hot queue stores the URLs with the topic word computer in their anchors or in the URLs
themselves. The second queue, url queue, keeps the rest of the URLs. The crawler always prefers
to take URLs to visit from hot queue.
Figure 12 shows the P
ST
results for this crawler for the IS(p) importance metric dened above.
The horizontal axis represents the fraction of the Stanford Web pages that has been crawled and the
vertical axis shows the crawled fraction of the total hot pages. The results show that the backlink-
count and the PageRank crawler behaved no better than a random crawler. Only the breadth-rst
crawler gave a reasonable result. This result is rather unexpected: All three crawlers dier only in
their ordering metrics, which are neutral to the page content. All crawlers visited computer-related
URLs immediately after their discovery. Therefore, all the schemes are theoretically equivalent and
should give comparable results.
The observed unexpected performance dierence arises from the breadth-rst crawlers FIFO
nature. The breadth-rst crawler fetches pages in the order they are found. If a computer-related
page is crawled earlier, then the crawler discovers and visits its child pages earlier as well. These
pages have a tendency to be computer related, so performance is better.
Thus the observed property is that if a page has a high IS(p) value, then its children are likely
to have a high IS(p) value too. To take advantage of this property, we modied our crawler as
shown in Figure 13. This crawler places in the hot queue URLs that have the target keyword in
their anchor or URL, or that are within 3 links from a hot page.
Figure 14 illustrates the result of this crawling strategy. All crawlers showed signicant im-
provement and the dierence between the breadth-rst crawler and the others decreased. While
the breadth-rst crawler is still superior to the other two, we believe that this dierence is mainly
4
In our third experiment we consider a dierent topic, admission, and show the results.
13
Algorithm 5.2 Crawling algorithm (modied similarity based)
Input: starting url: seed URL
Procedure:
[1] enqueue(url queue, starting url)
[2] while (not empty(hot queue) and not empty(url queue))
[3] url = dequeue2(hot queue, url queue)
[4] page = crawl page(url)
[5] enqueue(crawled pages, (url, page))
[6] url list = extract urls(page)
[7] foreach u in url list
[8] enqueue(links, (url, u))
[9] if (u / url queue and u / hot queue and (u,-)/ crawled pages)
[10] if (u contains computer in anchor or url)
[11] enqueue(hot queue, u)
[12] else
[13] enqueue(url queue, u)
[14] reorder queue(url queue)
[15] reorder queue(hot queue)
Function description:
dequeue2(queue1, queue2): if (not empty(queue1)) dequeue(queue1)
else dequeue(queue2)
Figure 11: Similarity-based crawling algorithm
0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
1
P
S
T
Ordering metric O(u)
Breadth
Random
Ideal
Backlink
PageRank
Figure 12: Basic similarity-based crawler. I(p) = IS(p); topic is computer.
14
Algorithm 5.3 Crawling algorithm (similarity based)
Input: starting url: seed URL
Procedure:
[1] enqueue(url queue, starting url)
[2] while (not empty(hot queue) and not empty(url queue))
[3] url = dequeue2(hot queue, url queue)
[4] page = crawl page(url)
[5] if (page contains 10 or more computer in body
or one computer in title)
[6] hot[url] = True
[7] enqueue(crawled pages, (url, page))
[8] url list = extract urls(page)
[9] foreach u in url list
[10] enqueue(links, (url, u))
[11] if (u / url queue and u / hot queue and (u,-)/ crawled pages)
[12] if (u contains computer in anchor or url)
[13] enqueue(hot queue, u)
[14] else if (distance from hotpage(u) < 3)
[15] enqueue(hot queue, u)
[16] else
[17] enqueue(url queue, u)
[18] reorder queue(url queue)
[19] reorder queue(hot queue)
Function description:
distance from hotpage(u):
return 0 if (hot[u] = True)
return 1 if (hot[v] = True and (v, u) links for some v)
. . .
Figure 13: Modied similarity-based crawling algorithm
15
0.2 0.4 0.6 0.8 1
Fraction of Stanford Web crawled
0.2
0.4
0.6
0.8
1
P
S
T
Ordering metric O(u)
Breadth
PageRank
Random
Ideal
Backlink
Figure 14: Modied similarity-based crawler. I(p) = IS(p); topic is computer.
0.2 0.4 0.6 0.8 1
Fraction of Stanford Web crawled
0.2
0.4
0.6
0.8
1
P
S
T
Ordering metric O(u)
Breadth
PageRank
Random
Ideal
Backlink
Figure 15: Modied similarity-based crawler. Topic is admission.
due to statistical variation. In our other experiments, including the next one, the PageRank crawler
shows similar or sometimes better performance than the breadth-rst crawler.
In our nal experiment, with results shown in Figure 15, we repeat the scenario reported in
Figure 14 with a dierent query topic. In this case, the word admission is considered of interest.
Details are not identical to the previous case, but the overall conclusion is the same: When similarity
is important, it is eective to use an ordering metric that considers 1) the content of anchors and
URLs and 2) the distance to the hot pages that have been discovered.
6 Conclusion
In this paper we addressed the problem of ordering URLs for crawling. We dened several dierent
kinds of importance metrics, and built three models to evaluate crawlers. We experimentally
evaluated several combinations of importance and ordering metrics, using the Stanford Web pages.
In general, our results show that PageRank, IR

(p), is an excellent ordering metric when either

pages with many backlinks or with high PageRank are sought. In addition, if the similarity to a
16
driving query is important, it is useful to visit earlier URLs that:
Have anchor text that is similar to the driving query;
Have some of the query terms within the URL itself; or
Have a short link distance to a page that is known to be hot.
With a good ordering strategy, we can build crawlers that can obtain a signicant portion of
the hot pages relatively early. This property can be extremely useful when we are trying to crawl
a fraction of the Web, when our resources are limited, or when we need to revisit pages often to
detect changes.
One limitation of our experiments is that they were run only over the Stanford Web pages. We
believe that the Stanford pages are a reasonable sample: For example, they are managed by many
dierent people who structure their pages in a variety of ways. They include many individual home
pages, and also many clusters that are carefully managed by organizations. Nevertheless, it will
be interesting to investigate in the future non-Stanford Web pages to analyze structural dierences
and their implication for crawling.
7 Acknowledgement
This material is based upon work supported by the National Science Foundation under Cooperative
Agreement IRI-9411306. Funding for this cooperative agreement is also provided by DARPA,
NASA, and the industrial partners of the Stanford Digital Libraries Project.
References
[BP98] Sergey Brin and Lawrence Page. The anatomy of a large-scale hypertextual web search
engine. In Proceedings of the Seventh International World-Wide Web Conference, Bris-
bane, Australia, April 1998.
[Goo] Google Inc. http://www.google.com.
[Kah97] Brewster Kahle. Archiving the internet. Scientic American, March 1997.
[Kos95] Martijn Koster. Robots in the web: threat or treat? ConneXions, 4(4), April 1995.
[Mar97] Massimo Marchiori. The quest for correct information on the web: Hyper search engines.
In Proceedings of the Sixth International World-Wide Web Conference, pages 265276,
Santa Clara, California, April 1997.
[PB98] Lawrence Page and Sergey Brin. The anatomy of a large-scale hypertextual web search
engine. In Proceedings of the Seventh International World-Wide Web Conference, Bris-
bane, Australia, April 1998.
[Pin94] Brian Pinkerton. Finding what people want: Experiences with the web crawler. In
Proceedings of the Second World-Wide Web Conference, Chicago, Illinois, October 1994.
17
[PPR96] Peter Pirolli, James Pitkow, and Ramana Rao. Silk from a sows ear: Extracting usable
structures from the web. In Proceedings of the Conference on Human Factors in Com-
puting Systems CHI96, pages 118125, Vancouver, British Columbia, Canada, April
1996.
[Rob] Robots exclusion protocol. http://info.webcrawler.com/mak/projects/robots/
exclusion.html.
[Sal83] Gerard Salton. Introduction to modern information retrieval. McGraw-Hill, rst edition,
1983.
[YLYL95] Budi Yuwono, Savio L. Lam, Jerry H. Ying, and Dik L. Lee. A world wide web re-
source discovery system. In Proceedings of the Fourth International World-Wide Web
Conference, Darmstadt, Germany, April 1995.
18