Library Hi Tech
Analysis of EZproxy server logs to visualise research activity
in Curtin’s online library
Journal:
Library Hi Tech
Manuscript ID
Draft
Manuscript Type:
Original Article
Keywords:
Data visualisation, EZproxy server logs, User-centered design, Collection
Management, Academic libraries, information resources management
Library Hi Tech
Library Hi Tech
1
Analysis of EZproxy server logs to visualise research activity
in Curtin’s online library
Introduction
Curtin Library has a substantial dataset of logged, authenticated use of its online library collection,
comprising databases, eJournals and eBooks dating from 2013. The EZproxy software writes about
30 million lines a month, and this rich dataset was accessible for this project.
Making sense of and drawing meanings from the raw data, which is mainly in the textual
format of URL codes, is nearly impossible. The Curtin library team reported an unsuccessful attempt
ten years ago, reporting that it was impossible to comprehend so much data using the human eye,
as it all looked similar, with no distinct observable trends in users’ information-seeking behaviours.
Scholarly findings about visualisations based on EZproxy data have been traditionally static,
as in the work by Bhaskar et al. (2014), Coombs (2005), Chan (2014), Grace and Bremner (2004),
Lewellen et al. (2016) and Sharman (2017). Dynamic visualisations are uncommon in this field, but
Archambault et al. (2015, p.1) argue that a framework for meaningful data visualisation has merit:
There are advantages to presenting data visually rat
her than as a set of flat
statistics. Proper data visualization facilitates the recognition of patterns and
relationships to communicate a message in a more compelling and interesting
way. It allows the complexity of the data to be understood more easily.
W
e agree with Archambault et al. (2015) that the creation of a set of visualisations that is
dynamic and immersive would offer friendlier interactions with the dataset, allowing one to look at
it in a way that normally is not possible. The creation of a 3D virtual space where one could move
around and view the data freely was envisioned to provide the library team with an immersive
encounter with the dataset and an opportunity to explore their users’ information seeking
behaviour.
An opportunity to work with the EZproxy dataset was made possible with technical and
financial support from the Curtin HIVE (Hub for Immersive Visualisation and eResearch) Internship
Program, which allows a Curtin student to undertake a ten-week, full-time investigation of the
application of visualisation technologies to a discipline area. Interns have regular access to the HIVE,
are supported by its expert staff and supervised by a library and information science discipline leader
and the Curtin Library team; the latter were the clients for this project.
In this paper we describe a project that aimed to visualise the EZproxy dataset to draw
inferences about Curtin library users’ information seeking behaviour and collection use in the
virtual/online environment. EZproxy offered a rich source of data for analysis, containing a detailed
log of HTTP requests processed through the library’s authentication servers. Another user identity
dataset was merged with EZproxy to identify each users’ profile: whether staff or student; their
geolocation data; and if their requests were processed in or out of Curtin’s IP subnet. Analysis of
these logs led to visualisations of hidden trends in users’ collection of information, providing insights
into where and when users accessed the virtual library, what information seeking activities they
performed, whether they browsed or downloaded full-text journal articles, and which online
databases they accessed frequently.
Visual awareness and understanding of these trends will help the library to make important
budget decisions about their collection subscriptions and assist with strategic planning about their
service delivery options (Grace and Bremner, 2004).
Page 1 of 24 Library Hi Tech
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Library Hi Tech
2
Background
Curtin University Library
Curtin Library supports the University’s learning, teaching and research requirements and supports
more than 58,000 students and 3,700 staff worldwide. Its online collection comprises at least
250,000 eBooks, 15,000 streamed videos, 150,000 journal subscriptions, 600 electronic databases
and 48,000 institutional repository records. The library’s acquisitions budget is over $10 million.
Access and authentication are administered by EZproxy, software licensed by OCLC. EZproxy reduces
the number of authorisations for users and ensures remote access to content is secure and complies
with licensing arrangements.
What is an EZproxy server?
EZproxy is an internet proxy server that is primarily used to first authenticate then allow computers
outside a library network to access content provided by the library without requiring additional
logins. Figure 1 illustrates how this works: when a user accesses the Curtin catalogue and attempts
to access content on an Online Database Vendor’s website, for example ProQuest, they are
authenticated by the library’s authentication system, which then creates a session with the EZproxy
server and redirects the user’s browser there, allowing access to all subscribed library content.
[Insert Figure 1: How the EZprozy server works?]
All hypertext transfer protocol (HTTP) requests that are sent through EZproxy are logged by
its server. For example, if a user is directed to a website through Google Scholar, once they are
authenticated by Curtin’s Authentication server, all requests for this website will be logged within
the dataset, creating a valuable set of information about who accesses resources via the EZproxy
server that can be utilised to visualise users’ information search and retrieval behaviours.
A user may access the library’s online databases from a personal browser by going to the
library catalogue or A – Z list of subscription databases, or through Google Scholar. When a user is
successfully authenticated, a cookie is sent to the browser by EZproxy. The browser then presents
this cookie for each access to EZproxy, allowing EZproxy to check the user’s access rights each time.
EZproxy handles access by URL rewriting: taking the URL requested by a user and modifying it so that
the web server holding the content accepts the request. This removes the need for the user to use a
separate login, as they are already authenticated by EZproxy. An example is when a user tries to
access the online database ‘proquest.com’: the user’s browser will be informed of the URL that
contains the cookie, ‘proquest.dbgq.lis.curtin.edu.au’ —everything after ProQuest is the cookie.
(‘URL Rewriting’, 2018)
How other Academic Institutions use the EZproxy Server and
Dataset
A literature review shows how other academic libraries handle their EZproxy data logs. Academic
libraries from the United Kingdom (Sharman, 2017; Grace and Bremner, 2004) and the United States
(Chan, 2014; Coombs, 2005; Lewellen et al., 2016) report their experiences working with EZproxy,
how their findings assisted in improving their service, and understanding of the usage patterns of
their electronic resources. The Open University library reported using their EZproxy server data to
evaluate their electronic resource subscriptions comprehensively and monitor resource usage trends
by staff and students; the findings enabled them to measure the performance of their library
services, and led to the development of performance indicators to measure their service quality and
Page 2 of 24Library Hi Tech
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Library Hi Tech
3
patron impact (Grace and Bremner, 2004, p. 164). Sharman (2017) reports how the University of
Huddersfield combined their EZproxy dataset with book loans from the library management system
and statistics from library visits; they found correlations between the low use of library resources
and final grades among Chinese students compared to their UK peers.
In the United States, Coombs (2005) reports how the SUNY Cortland University used the
EZproxy dataset to gain insights into online collection use. Chan (2014) discusses how student course
enrolment data was merged with the EZproxy dataset at California State University to develop
personalised e-library services. Lewellen et al. (2016) describe how the University of Massachusetts
Amherst used the dataset to investigate the use of e-books compared to print books.
These studies demonstrate that it is possible to extract useful information about trends and
usage patterns from EZproxy datasets, and that these can be used to improve library services in
universities. However, working with the complex raw EZproxy dataset is neither easy nor user-
friendly, as revealed by the East Kentucky University’s experience. This university analysed its
EZproxy dataset to determine use patterns because this information was not provided by the online
database vendors. Hence, the university firstly used an OpenURL link resolver, which provides a set
of reports that enable robust analysis of user behaviour, as shown in Figure 2.
[Insert Figure 2. An example of OpenURL results (adapted from Smith and Arneson, 2017, figure 3)]
This snapshot of a report generated from the EZproxy dataset is detailed, but the textual
data is visually unappealing—in fact, it looks just how our project’s EZproxy data displayed when
exported into MS Excel.
The second method that East Kentucky University used to extract data from their EZproxy
data logs was by using the command-line grep tool, as shown in Figure 3. The grep tool is a Unix
based pattern matcher that searches through plain text data sets in the target file and outputs all
lines that match a regular expression of these patterns. Both the construction of these patterns and
working with the grep utility tool are daunting tasks for novices.
[Insert Figure 3. Example of grep on a small sample file
(‘Pipe, Grep and Sort Command in Linux/Unix with Examples’, 2018)]
There are three disadvantages to using this method. First, the output of this tool is in the
same format that it goes in, as it does not change the lines; instead, it writes the files that match the
pattern to a new file. Second, the single grep command-line will not work for all queries as it cannot
cope with the varying file formats used by different database vendors. Third, it does not enable the
development of data visualisations, as its functionalities are limited to searching the file for certain
queries.
Given our research aim to find out about Curtin’s online collections usage trends, an
understanding via the literature review on how others achieved this was of interest. Hence, Morton-
Owens and Hanson’s (2012) research about how they analysed EZproxy logs through the creation of
dashboard charts based on EZproxy data was useful. They report running a few basic calculations to
highlight significant changes within their data, which produced two charts indicating trends in
resource use and variations in use of specific resources. It was our understanding that often
additional analytical software is required to make sense of this complex EZproxy data. Thus, it was
useful to read that Austin College used Google Analytics on their EZproxy dataset to identify which
were the main resources being used, when (time/day) the most use took place each week, and what
devices were being used to access their library (Ajamie et al., 2014).
EZproxy server logs are a rich source offering immense opportunities for harnessing their
data to tell stories of how library users search for information and how the library’s information
Page 3 of 24 Library Hi Tech
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Library Hi Tech
4
sources are used at any time, or 24/7/365. Presenting this information visually will assist both in
planning and management of the library collection and in improving its service delivery (Grace and
Bremner, 2004; Kay, 2014). Knowing how students use library resources can aid in focusing on
teaching those information literacy skills most needed; and to identify groups of students needing
specific remedial skill teaching (Kay, 2014). It can also enable the delivery of e-library services, such
as the personalisation of library websites/portals for students and staff in specific discipline areas to
provide rapid access to their most frequently used online resources (Chan, 2013; Grace and
Bremner, 2004, p. 162;).
This review identifies a gap in the work about using data visualisation tools and techniques
to interrogate the rich but complex EZproxy dataset. At the moment, all data is presented as static
flat graphs and tables. No-one so far has developed an interactive search interface that is friendly to
library decision-makers, who are not trained to work with or understand complex computing
datasets. There is also a lack of built-in user-friendly search interfaces that offer library decision-
makers the opportunity to explore visually their clients’ techniques of information-seeking. In the
absence of these functionalities, it is not yet possible for library professionals to have an immersive
experience that enables them better to understand their clients’ usage patterns. Hence, it is
worthwhile to offer librarians both a familiar and an ease to use search interface that is similar to
their online library catalogue interfaces to search for usage information about their online resources.
Further, these data visualisation tools will provide a visually immersive experience whereby they can
intuitively explore the dataset to identify usage patterns and statistics. This project offers a new and
friendlier way for librarians to analyse datasets that have traditionally been difficult for them to
understand.
Definition of information seeking behaviour
Given that the EZproxy server logs staff and student interactions with various online databases and
electronic resources provided by the library. It is necessary to define the broad behaviours the
findings from the EZproxy datasets reveal. The information seeking behaviour literature is vast and
only key definitions that are relevant for this project are described in our article. We adopt Joseph et
al.’s (2013) definition: that seeking information ‘relates to the process of identifying an information
need, then sourcing and accessing the necessary information avenues to address that need’.
We consider information retrieval is a crucial activity in the information-seeking process,
especially in the context of this research: that users wish to fulfil their informational needs from the
widest possible pool of electronic resources. Meadow et al.’s (2007, p. 2) definition that information
retrieval ‘involves finding some desired information in a store of information or a database’ is
applicable here. Information ‘search behaviour’ also needs to be defined. Joseph et al. (2013) state
that information search behaviour comprises ‘search processes’ and ‘search activities’. Search
processes involve several sequential but iterative stages of judgement, option selection and
decision-making (Ellis 2005; Henefer and Fulton 2005; Kuhlthau 1988; Leckie et al. 1996; Marchionini
and White 2007; Meho and Tibbo 2003). Joseph et al. (2013, p. 6) state:
Search activities refer to the actions users enact d
uring the iterative process of
moving the information search from start to closure. These actions include
browsing, navigating or extracting information. Both information search
processes and search activities comprise the search behaviour of users of these
systems.
These simple and practical definitions are useful gu
ides to our discussions on the
research aims of our project described next.
Page 4 of 24Library Hi Tech
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60