GLOBAL SEARCH TOOL FOR THE ADVANCED PHOTON SOURCE
INTEGRATED RELATIONAL MODEL OF INSTALLED SYSTEMS
(IRMIS) DATABASE *
D. E. R. Quock
#
, ANL, Argonne, IL 60439, U.S.A.,
M. B. Cianciarulo, Purdue University, West Lafayette, IN 46324, U.S.A.
Abstract
The Integrated Relational Model of Installed Systems
(IRMIS) is a relational database tool that has been
implemented at the Advanced Photon Source to maintain
an updated account of approximately 600 control system
software applications, 400,000 process variables, and
30,000 control system hardware components. To
effectively display this large amount of control system
information to operators and engineers, IRMIS was
initially built with nine Web-based viewers: Applications
Organizing Index, IOC, PLC, Component Type, Installed
Components, Network, Controls Spares, Process
Variables, and Cables. However, since each viewer is
designed to provide details from only one major category
of the control system, the necessity for a one-stop global
search tool for the entire database became apparent. The
user requirements for extremely fast database search time
and ease of navigation through search results led to the
choice of Asynchronous JavaScript and XML (AJAX)
technology in the implementation of the IRMIS global
search tool. Unique features of the global search tool
include a two-tier level of displayed search results, and a
database data integrity validation and reporting
mechanism.
HISTORY OF IRMIS
Don Dohan, while working at the Argonne National
Laboratory (ANL) Advanced Photon Source (APS), set to
the task of developing a software framework for modeling
the EPICS installed control software. He used a relational
database as the backbone for describing how EPICS
process variable attributes are linked to each other [1].
Subsequently, Ned Arnold, of ANL, and Don partnered
together to expand the database schema to include
installed control system hardware in a connection-
oriented approach [2]. This relational database application
became what is now known as IRMIS, and
implementations spawned from IRMIS are currently in
use at GANIL, the Canadian Light Source, TRIUMF,
SLAC, BESSY, SNS, and PSI. Over time, and with the
addition of more IRMIS developers, the IRMIS Web-
based user interface at APS evolved to include nine
separate “viewers” for displaying categorized, detailed
information about control system hardware and software
(see Figure 1).
Figure 1: APS IRMIS home page.
GLOBAL SEARCH TOOL
A Global Search Tool (GST) for locating a string of text
in any category of IRMIS data has recently been added to
the suite of IRMIS viewers. AJAX technology was
selected for building a highly responsive Web-based user
interface. The motivation behind creating this Google-like
keyword search tool for IRMIS is as follows:
x The IRMIS relational database at APS comprises
seventy tables. An organized and logical means for
quickly searching through this expansive amount of
information when only a small amount of
information is initially known by the user, such as a
single word or abbreviation, was highly desired to
aid engineers in rapidly troubleshooting real-time
control system issues.
x IRMIS has six viewers written in PHP language. This
type of Web page display requires complete
reloading each time new data is retrieved from the
IRMIS database. This makes for a sluggish user
interface environment.
x The other three viewers are written in Java with
Hibernate object relational database mapping
software. The six PHP and three Java viewers are
currently not connected to each other through
software links, and for the most part, the PHP
viewers are not interlinked – the user must use the
browser back button and start over on their search of
IRMIS data to get to other categories of control
system software or hardware.
___________________________________________
*Work supported by the U.S. Department of Energy, Office of Science,
Office of Basic Energy Sciences, under Contract No.
DE-AC02-06CH11357.
#
x The Java/Hibernate viewers are relatively slow when
downloading the initial set of values retrieved from
the IRMIS database, and once loaded, the
Proceedings of ICALEPCS07, Knoxville, Tennessee, USA RPPB34
Operational Tools
683
information is static during the entire viewing
session.
Software Architecture
The Global Search Tool is built with AJAX technology
for the Web-based user interface, PHP for the server side
processing of user requests, and MySQL as the IRMIS
relational database. AJAX is not a specific software
language, but rather a software design concept on how to
combine JavaScript, HTML, Cascading Style Sheets
(CSS), and Extensible Markup Language (XML) to make
asynchronous calls to the server via the JavaScript
XMLHttpRequest object. The XMLHttpRequest object is
the key component of AJAX that provides the ability to
refresh a subsection of a Web page with new data
retrieved from a server while the remainder of the Web
page remains intact and available for other user
interactions. XML is the markup tag syntax used for
transferring data from/to JavaScript running on the client
(user’s Web page) to/from PHP running on the Web
server.
JavaScript ÅÆ XML ÅÆ PHP ÅÆ MySQL
Many excellent books for the novice have been published
on AJAX [3].
User Interface
The Global Search Tool provides three levels of detail
for defining the user’s search criteria. All of IRMIS can
be searched simply by typing in a string of text. One or
more major categories of IRMIS can be selected for
narrowing the search. The major categories currently
defined include AOI, IOC, PLC, Component Type,
Installed Components, and Process Variables. A user’s
search can optionally be further refined by selecting only
from the minor categories of IRMIS. Each major
category is broken down into its constituent minor
categories as shown in Figure 2. For example, the major
category IOC has minor categories: IOC Name, System,
Location, Cognizant Developer, and Cognizant
Technician.
Figure 2: GST major and minor search categories.
The asynchronous behavior of the Global Search Tool
is designed such that a new search of the IRMIS database
is performed in response to every user keystroke in the
text entry box. A minimum of three characters is required
in the search string before a request is sent from the client
to the server. A new search is not requested by the client
in the case where: the current search string is over three
characters in length and the current search resulted in zero
matches -- the user then shortens the search string by
deleting characters. This logic is used to not waste time
processing a non-productive search request.
The search results are presented in two levels of detail.
The first level of detail provided to the user shows the
categories searched on and the number of matched results
for each category. If the number is greater than zero, then
that number is displayed as a link to launch a new tabbed
Web page with further details of that specific category. If
the user selects this link, the GST then submits another
request to the server to retrieve a general list of the initial
set of matched results. Thus, the GST retrieves more
extensive information only as the user requests it.
The newly created tabbed Web page includes a column
of checkboxes for each match in the search. As a
checkbox is selected, the JavaScript client issues yet
another query request to the PHP server-side application,
this time for detailed information on each matched result,
and a table is created to the right of the Web page with a
new row for each check box selection. The example
shown in Figure 3 is for a search done on the string
‘safety’ in major category AOI. The matched search
string results are highlighted with a yellow background
and bold font.
Figure 3: GST detailed search results.
Other features of the Global Search Tool include:
x EPICS process variables are not included in the
default top level search of all major categories due to
the large number of process variables in the IRMIS
database (over 400,000) and the delayed search time
required for this category. The Advanced Search
RPPB34 Proceedings of ICALEPCS07, Knoxville, Tennessee, USA
Operational Tools
684
button must be used to reach the option to search
process variables.
x If there is a discrepancy between the initial data
search and the general list data search for a given
category, this is an indication of missing data in the
IRMIS database for related attributes of a category.
The GST writes information on this incomplete
search result to a log file for later IRMIS data
inspection.
x Links to existing IRMIS PHP viewers are provided in
the GST table of search results details.
x A report tool is provided to capture the search results
to a printer-friendly Web page, a comma separated
value (CSV) file such as Microsoft Excel, or to a text
file.
x Helpful hints for using the GST can be viewed
interactively on mouse-over action of a text link.
x The GST was built in a generic software framework
with JavaScript objects and PHP functions so that it
can be easily extended as needed for new major and
minor categories.
Performance
One of the main user requirements for the IRMIS
Global Search Tool is a quick response time. Due to the
large volume of data available in IRMIS, this presented
interesting challenges in the use of AJAX technology.
Rather than following the typical Model-View-Controller
(MVC) philosophy of decoupling data (model) from the
user interface (view), a different approach was adopted in
the GST. To speed up the response time to the user upon
lengthy retrievals and sorting of a large number of results,
the decision was made to add HTML code for checkboxes
on the server side where PHP sorting of data returned
from the MySQL database is performed. The PHP creates
a string of HTML checkbox code that is embedded along
with the actual IRMIS database queried results in XML
elements. The XML elements are then returned to the
client side JavaScript application as a single text sting and
are subsequently parsed as a JavaScript responseXML
document type. The JavaScript client does not need to
further sort the returned data as the code for HTML
checkboxes has already been added on the server side.
The time savings in performing a sort on the IRMIS
database queried results only once (in PHP server side)
was significant – from several minutes to almost
instantaneous updates to the user Web page.
Currently, typical response times for searches in the
GST are:
x Process Variables: highest return of two minutes for
detailed information on 7,794 matches
x AOIs: approximately 30 seconds to return detailed
information on 681 matches
x Installed Components: approximately four seconds to
return detailed information on 109 matches
Development Obstacles and Design
Shortcomings
AJAX technology comes with its own set of
shortcomings as a Web application. JavaScript and CSS
are inherently highly susceptible and dependent on the
browser type (e.g., Firefox or Internet Explorer). A lot of
time and frustration was spent on developing code that
was compatible among both browsers. The asynchronous
nature of AJAX makes it difficult to stop client-side
requests to the server once they have been sent. Also,
since AJAX-enabled Web pages are dynamically enabled,
the browser “back button” can produce unexpected
results. A design shortcoming of the Global Search Tool is
that new data from the server are requested from the client
(pulled) instead of pushed from the server as data in the
MySQL database changes. This means that the user must
explicitly request updates by either making a new search
request through the Global Search Tool or by refreshing
the Web page.
Future Enhancements
Although the outcome of the Global Search Tool was
successful, there remains much room for improvement
and added features. Future planned upgrades include
limiting the number of process variables that can be
viewed at a time to speed up response time, changing the
log file display format to XML and displaying as a Web
page using Extensible Stylesheet Language
Transformations (XSLT), and adding more minor
categories to the major category EPICS Record.
CONCLUSION
The use of AJAX technology for creating a
comprehensive search tool for the IRMIS database proved
to be extremely beneficial. The rapid response time for
querying the IRMIS database and displaying results back
to the user upon every user keystroke is impressive. This
project served as a useful prototype for deploying future
controls Web-based applications in AJAX.
REFERENCES
[1] D. Dohan, “Decision Support Facility for the APS
Control System,” ICALEPCS 2001, 271 (2001);
http://www.jacow.org.
[2] D. Dohan, N. Arnold, “Connection-Oriented
Relational Database of the APS Control System
Hardware,” PAC 2003, 2327 (2003);
http://www.jacow.org.
[3] T. Negrino and D. Smith, JavaScript and AJAX for
the Web, Sixth Edition, Peachpit Press, Berkeley, CA
2007.
Proceedings of ICALEPCS07, Knoxville, Tennessee, USA RPPB34
Operational Tools
685
