=Paper=
{{Paper
|id=None
|storemode=property
|title=Recipes 2.0: Building for Today and Tomorrow
|pdfUrl=https://ceur-ws.org/Vol-993/paper16.pdf
|volume=Vol-993
|dblpUrl=https://dblp.org/rec/conf/iwsg/DooleyH13
}}
==Recipes 2.0: Building for Today and Tomorrow==
https://ceur-ws.org/Vol-993/paper16.pdf
Recipes 2.0: Building for Today and Tomorrow
Rion Dooley Matthew R. Hanlon
Texas Advanced Computing Center Texas Advanced Computing Center
The University of Texas at Austin The University of Texas at Austin
Austin, US Austin, US
Abstract—The history of science gateway development has, in multiple investments of time writing system-specific code
many ways, been a story of the “Haves” vs. the “Have-nots.” when the web only requires a single investment.
Large infrastructure projects led the way, building thick client
portals to provide coherent interfaces to an incoherent Despite the cost and complexity, projects that could afford
environment. Contrast this with the way the modern web is to make the investment in a portal did so gladly because the
designed using light, front end components and outsourcing end product was well worth the cost. Portals brought cohesion
much of the heavy lifting to a mash-up of REST APIs, and it is to complicated infrastructure and made computational science
easy to see why modern web applications can be prototyped and accessible to researchers without computer science degrees.
refined into stable products in the time it previously took thick They pulled the focus away from the machines and put it back
client portals to do an initial release. This paper argues that a onto the science.
“build for today” philosophy can lead to the rapid development
of science gateways to serve the “Have-nots.” Also presented is a Portals such as Cipres [1], GridChem [2], UltraScan [3],
set of responsive front end components built on top of the iPlant Galaxy [4], and NanoHub [5], just to name a few examples in
Foundation API that provide the boilerplate for rapid the United States, continue to provide tremendous value to
development of lightweight science gateways using only HTML, their user communities. The researchers using these portals
JavaScript, and CSS. Using these components, developers can have made discoveries leading to hundreds of published
easily stand up new gateways or quickly add new functionality to papers, multiple thesis and dissertations, and insights that
existing ones. would have taken significantly longer to realize, if at all. One
cannot deny the value of such portals in today’s scientific
Keywords— Science Gateway, REST, API, web service, process.
AGAVE, HTML5, JavaScript, web
The challenge portal-driven science faces is that for every
scientist that has a portal like Cipres at their disposal there are
I. INTRODUCTION
hundreds more in the same domain who do not. No portal can
The history of science gateway development has, in many meet the needs of everyone. Successful portals find their niche
ways, been a story of the “Haves” versus the “Have-nots.” and focus on providing value to the researchers in that niche.
Large infrastructure projects led the way, building thick client Inherent in the design of a successful portal is the realization
portals to piece together incongruent service stacks and provide that it cannot and will not meet the needs of the vast majority
cohesion to an incoherent environment. The field was of scientists who could otherwise derive value from similar
dominated so thoroughly by these heavyweight portals, that the tools. Thus, even within the highly technical research
terms portal and gateway became interchangeable. A gateway landscape there is a digital divide [6] between those who have
was no longer just a means of access, it was an ecosystem of advanced portal technology to facilitate their work and those
moving parts that all had to be managed and maintained over who do not.
time for the gateway to work. The concept of modular design
became a relative term. If one could take a component out of Exact numbers are difficult to obtain, but a rough
one monolithic framework instance and add it to another approximation is possible. The National Science Foundation’s
monolithic framework instance, then the component was Science and Engineering Indicator Report for 2012 (SEI) states
modular. Disregard the background processes, supporting that as of 2010, the US Science and Engineering (S&E)
services, and database that needed replication in order for the workforce is 6.65 million people [7]. Of them, 31% describe
module to work. If the UI could be reused, the component was research and development (R&D) as a major work activity. If
considered modular. we consider only those with doctorates, 12% of those who
describe R&D as a major work activity remain. This indicates
The resources required to build and maintain such portals that there are at least 247,000 PhD level workers in S&E
made finding portals with long-term success rare. Whereas at actively conducting research in the US. Add to this the
one time portals were built as thick desktop clients, one of the estimated 100,000 medical researchers in the US according to
reasons that portals gravitated from the desktop to the web was the Bureau of Labor and Statistics and we come to a lower
the ongoing cost of maintaining software on multiple operating bound of 347,000 for the number of researchers who could be
systems. Even applications written entirely in Java require impacted by portal technologies [8].
some platform-specific attention. That means multiple sets of
unit tests, multiple testing environments, and most importantly, Looking again at SEI, we see a reasonably proportional
investment in R&D across US S&E companies of roughly 6%.
Given that the Pareto Principal applies to revenue distribution release of GridChem provided federated identity management,
among businesses, we can infer an 80/20 split among industrial job tracking, system monitoring, scheduling, enforcement of
researchers [9]. With 20% having access to the latest high proprietary software license agreements, distributed account
technology tools to perform their research and 80% utilizing management, large data management, full experiment
effective, but cost-restricted technologies. In academia, SEI reproducibility, and integration with application codes
shows that the top 100 spending universities spent 80% of the installed on the user’s local system. Many of the features took
academic R&D money in the US. This is significantly more a significant amount of time to build which pushed back the
lopsided ratio, but as a lower bound, the Pareto Principal holds first release of the software by nearly a year. However, after
for academic research as well. Thus, it is reasonable to assume
its first 3 years in production GridChem had enabled 500 plus
at least an 80/20 split between the haves and the have-not
researchers to publish over 60 papers and complete 6
across US R&D in both sectors today, indicating there are at
least 277,000 underserved researchers in the US alone. dissertations. The software was used as a teaching tool in
undergraduate chemistry classes at The Ohio State University,
the University of Illinois, and the University of Kentucky to
II. BUILDING FOR TOMORROW expose hundreds of students each semester to computational
How does one go about reaching the 277K scientists on the chemistry. The value of GridChem is obvious, however that
other side of the digital divide? Raising taxes to build 10,000 value came at the up front cost of 6 man-years of development
portals is not realistic. It also does not address the at a cost of $2.7M to provide enough features to
fundamentally deeper issue of utility. That portals provide simultaneously support undergraduate students and full
value to their users is well documented [10][11][12]. What professors alike. Further operation led to another $1M in
value they provide and at what cost are less well-documented funding to support workflow integration and expanded support
questions. We look at 5 portals from the XSEDE Gateways for determining appropriate parameters for use in different
Program [13] as short case studies. experiments.
Galaxy is an open, web-based platform for data intensive The Cipres Science Gateway is a public resource for
biomedical research. Scientists can download a copy of inference of large phylogenetic trees. As of this writing,
Galaxy for private use or they can use the hosted Galaxy Cipres exposes 30 different tools for use on a preconfigured
instance, often called Galaxy Main. The Galaxy Main portal set of systems ranging from large shared compute clusters to
contains over 2500 application codes in its “Shed” that users private virtual machines. Users access these tools through a
can leverage for their work. Historically, the vast majority of form-driven web interface. The process of developing Cipres
users select a small number of codes that they use for all their included building multiple interfaces for each applications, job
work. In 2012, users ran over 100k jobs a month through scheduling heuristics, data management, accounting systems,
Galaxy Main. In addition to application registration and job identity management, and integration with multiple
submission, Galaxy also supports visualization and data infrastructure providers. These features took a significant
publication. Both are popular features, but neither is the amount of time, $4.5M in funding from NSF, and a very
primary focus of the portal. Does that mean that they were a talented team of programmers to develop. The result of that
waste of time? No. Galaxy Main serves over 28k users. There work was a wildly successful portal. Cipres now serves over
are many other features built into Galaxy, but the point of this 700 users and has been used to run nearly 100k simulations
observation is that as a portal, Galaxy casts a wide net and burning over 15M compute hours. After 18 months in
tries to provide something of value to every one its users. The production, Cipres’ usage was outgrowing its infrastructure.
price of doing so is added complexity, greater development Due to the heavyweight nature of the infrastructure it took
costs, and a larger investment in supporting infrastructure to another year of development and $1.5M in funding from NSF
run the application. Initially funded by two awards totaling to allow them to scale out to other systems and move away
just under $1.4M in 2006 from the National Science from a community account model. While growth is a common
Foundation (NSF), the additional use cases necessitated problem of success, this particular problem came at the end of
another round of funding totaling $1.1M from NSF. Recently, the project’s original funding. Had it not been for the talent
to support the expanding user community and support and passion of the development team, Cipres would not have
different resource utilization patterns, another round of been able to address its growing pains and, as such, would
funding totally $5.8M was obtained from the National Institute have stalled until the next round of funding arrived.
of Health to carry the project through 2018. Even for a
successful portal like Galaxy Main, maintaining continuous NanoHub is web application built upon the Joomla CMS
funding and retaining talent are ongoing concerns. [14] and designed to support nanotechnology research and
education. It provides over 270 simulation tools, 3800
GridChem is a desktop application supporting the seminars, tutorials, and teaching materials, 200 distinct user
computational chemistry community. Its mission is to enable groups, and a mature workflow engine called Pegasus, which
computational and experimental scientists to do more supports job execution across heterogeneous systems. Behind
computational chemistry by providing capability computing NanoHub lies a series of web services, command line tools, a
resources and services at their fingertips. To that end, the first full CMS, and an application-authoring tool. The portal as a
whole was built to support a large community and it does so The development of the XUP is a continuation of the previous
very well. In 2010, NanoHub saw 10k users run 380k 5 years of development of the TeraGrid User Portal [18]. The
simulations. In 2011, 11k users ran 400k simulations. In 2012, initial cost of development for the first TeraGrid User Portal
12k users ran 410k simulations. Clearly a lot of people are was on the order of $800k. Since then another $1.7M has been
doing a lot of work and the growth is cumulative year over invested in the dedicated team of developers maintaining
year. Such usage indicates that the portal is reaching a active development on XUP, adding features, addressing user
significant number of people, exposing them to some issues, and providing the general maintenance required of a
functionality, helping them accomplish a specific task, and a portal with over 12,000 registered users that supports a variety
percentage are coming back year after year. The numbers are of user communities within the XSEDE organization, each
impressive, but the behavior is consistent with other portals. with different needs.
Users come in, find a few tools and/or features of value to
them, and make a routine using those specific tools and/or In order to provide all of this functionality, the XUP, and
features for the duration of their interaction with the portal. to a lesser extend the XSEDE website, rely on a suite of
services that provide the backend information services. These
Success comes at a price, and the price of building services include relational databases, non-relational “NoSQL”
NanoHub was $14M from NSF. Sustaining NanoHub amid databases, SOAP and REST web services, flat file parsers, and
rapid growth has been an even more expensive activity. Their many other services that interact directly with the resources in
latest round of funding is $21.9M from NSF starting in 2013. the XSEDE CI. The front end is built from many custom
To put that in perspective, NanoHub is a Joomla instance with developed and specialized portlet applications, as well as out-
a lot of custom plugins and some back-end services to support of-the-box Liferay portlets. The system works because the
running nanotechnology simulations at an average rate of one development team has administrative access to the entire
simulation every 78 seconds. Looking at the CMS alone, the XSEDE infrastructure. They are able to obtain information
site receives 8.5 million hits a month. That is roughly half the that other gateways simply do not have access to. As a result,
traffic of edublogs.com, the leading educational blog provider the portlets developed for XUP and the functionality they
in the world with 1.6 million blogs since 2005 [15]. Given provide cannot easily be replicated simply by copying over the
comparable expenditures and team sizes between the two portlet code.
organizations, the cost of custom development and supporting
the back-end infrastructure of NanoHub costs roughly 200% Each of the above portals is different in focus and function,
more than the total cost of running the website alone. but they are all successful science gateway projects and
provide broad functionality. That functionality is often
The Extreme Science and Engineering Discovery targeted at a small set of users who, for a given portal, will
Environment (XSEDE) [16] is a National Science Foundation only ever use a subset of the features. The cost of these portals
(NSF) funded national cyberinfrastructure (CI) that provides a in terms of time and effort are all measured in multiple man-
set of large resources for scientific simulation and analysis. years and millions of dollars before they ever had a single
The XSEDE User Portal (XUP), led by TACC, is the primary user. They were designed to accommodate thousands of users
interface for users to XSEDE. It provides user account when they went live and they made sure they could support a
management, project management, documentation, data thousand users before they tried to support one. From their
management, and a myriad of other features to help users be inception they were targeting long-term operational goals
productive on the XSEDE CI. It was built on the Liferay rather than short-term results. To be clear, there is nothing
Portal platform [17], an enterprise open-source Java portal wrong with that, but it is an important distinction to make. The
framework. The Liferay platform itself provides many features image of a successful science gateway promoted over the last
out of the box, including a content management system, wikis, decade was a portal built to support users of tomorrow rather
calendaring, web forms, user forums, security and access than something that will get the results they need today.
control, and user notifications. Liferay also provides a plugin
development platform for extending the portal with plugins III. BUILDING FOR TODAY
and portlets. As an enterprise portal, Liferay is one of the The reality for many scientists on the wrong side of the
leaders in the field, but there is a significant financial and digital divide is that they do not need portals built for
human cost associated with its use. The cost of training, tomorrow; they need gateways built for today. They are
professional consulting, and enterprise support must be content using their current workflows, but are willing to adopt
considered. technologies that make their workflows more efficient, more
powerful, or less painful. They will gladly set down Outlook
XUP has a very different focus than the previously for Gmail, their departmental FTP server for Dropbox, and the
mentioned science gateways, and it is more of a “destination server under their desk for a virtual machine on Amazon.
portal” then “science gateway”. But it is another example of a These scientists are not pushing the boundaries of size and
large, enterprise project that is designed to be a one-stop shop scale, but they are, in aggregate, performing the bulk of the
that provides users of XSEDE everything they need to be science done today.
productive on XSEDE, excepting streamlined job execution.
These scientists do not live in an enterprise world and memory, and disk available than the virtual machines
their experimental processes are much less rigid than those of powering the hosted services we rely upon. Furthermore,
the organizations building the previously mentioned portals modern web browsers are constantly evolving with powerful
above. These scientists look for silver bullets, or the next best new features both for the user and the developers of web
thing, to accelerate the time between proposal and discovery. applications. At the same time, as more browsers have adopted
And if a miracle doesn’t come, simply squeezing an extra 5% web standards put forth by the W3C [23] these features are
out of their week would be a huge win for them. more available for use natively in the browser without the
need for polyfills such as Adobe Flash [24]. Some of these
Whether they realize it or not, these scientists have features can even leverage advanced capabilities of the
embraced the spirit of Agile [19] development that drives underlying system such as GPU accelerated CSS rendering
today’s web ecosystem. In contrast to the monolithic and animations. The latest CSS modules, such as transforms
deliverable approach historically taken by portal projects, [25] and transitions [26] are even beginning to push the
today's web creates and innovates at a blazing pace. Working boundaries of 3D graphics. Combined, this makes the
from incremental release to incremental release, actively development of feature-rich, high-performance, and reliable
engaging users, and obsessing over a results-first focus web applications using only HTML, CSS, and JavaScript a
enables high quality sites and services to be created and reality.
refined into stable products in the time it takes most portals to
make their first release. One notable example being an By moving away from monolithic frameworks and large,
application called Burbn, which over a seven-month server-side stacks to client-side applications built using only
timeframe morphed from a web application to an iPhone app HTML, CSS, and JavaScript and leveraging RESTful APIs,
to a cross-platform application, then changed its focus and one can rapidly develop powerful, targeted applications that
relaunched as Instagram [20]. A second example is the social can be quickly deployed, are highly scalable and “cloud-
bulletin board site Pinterest, which spent 3 months in friendly.”
development before launch, then constantly adapted to user
feedback over the next year before expanding as an iPhone One example is a tool created by Andre Mercer, an
app and exploding into the giant of today [21]. A third undergraduate student at the University of Arizona. Andre
example is a relatively new startup called GivePulse [22], created a simple web page that submitted a request to the
which spent 4 months iterating over designs and features with iPlant Foundation API to run a GeneSeqer job [27]. He spent
local philanthropic organizations before publicly launching as an afternoon creating the page, then showed it to his
a site enabling the promotion, matchmaking, and coordination supervisor, iterated a handful of times on the wording and
of volunteers with events. While each of these examples gives default settings, then pushed it out into the group’s website.
launch timelines in terms of months, their feature development Jon Duvick, a bioinformatician in a sister group saw the tool
cycles were on the order of 1-2 weeks with updates and bug and decided to add it to his site as well as embedding it as part
fixes pushed out daily. of his cloud-based annotation pipeline. Based on the success
of the original tool, Andre is now adding data browsing via a
In each of these examples, the product that went to market jQuery [28] dialog box to the form so users can run analysis
was markedly different from the project that was originally on files stored in the Cloud as well their desktop.
conceived. They survived due to their ability to leverage
existing open source technologies, prototype ideas, and add IV. BUILDING ON A SOLID FOUNDATION
small bits of functionality that they could present to their One of the reasons that applications can be built with such
audience and find out if it had enough promise to invest more light front ends is that they now rely upon a growing number
time into its continued development. of web-friendly APIs for much of the work. The API
watchdog Programmable Web has tracked the growth and
When attempting to serve the needs of the lower half of adoption of APIs since 2005 and has seen an explosion of new
the digital divide, developers would do well to learn from APIs in the last 2 years [29]. Much of this growth has been
Instagram, Pinterest, and GivePulse and take these lessons to attributed to the fact that, “APIs are helping companies do
heart. Start first by understanding that not every project can or business, with the tradeoff between adding an external
should be the next big thing. Providing a tool that helps a dependency being out-shined by the ability to move faster
researcher to see a problem in a different light and enables the building upon someone else’s expertise [30].” In short, APIs
discovery of a solution is a significant contribution in its own allow companies to run lighter and move faster.
right. The gateway does not need to serve every conceivable
user community to be a success. For new applications the abundance of APIs completely
changes the established paradigm. API providers offering
While much of what one interacts with on the web is access to cloud storage, authentication, identity management,
provided as a hosted service, i.e. Facebook, Gmail, DailyMile, and Backend-as-a-Service (BaaS) [31] have redefined how
etc., there is no reason that every gateway should be a hosted applications are built. Things that used to take months to build
service. Most new desktop computers have more CPUs, and test are now leveraged as hosted services and integrated in
an afternoon. One well-known benefactor of building on the • Data: Acts as a Rosetta stone for biological data.
shoulders of other APIs is the communication platform Supports the conversion of data between known
provider Twilio [32]. From its inception Twilio has leveraged formats.
Amazon Web Services to handle spikes in demand and offload • IO: provides multiprotocol data movement and
much of its compute load while focusing on the core part of management.
their service, the communication platform. • Jobs: Handles the end-to-end execution of registered
applications on a heterogeneous set of systems
Of the thousands of public APIs available today, and the ranging from HPC to raw VMs.
hundreds targeted towards science, there are remarkably few • Monitor: constantly monitors Foundation and its
that provide a generic platform for computational science. The dependent services. Provides real-time and historical
SoapLab [33] project provides mechanisms for accessing monitoring test results.
SOAP services through a common interface, but it does not • PostIt: pre-authenticated URL shortening.
deal with federated identity, sharing, or access control. The • Profile: search and view profiles of other users within
NEWT project exposes HPC systems on the web, but is the API.
restricted in scope to systems and services at NERSC [34].
• Systems: provides information about systems
Recently, the CHAIN project has promoted an end-to-end
available from Foundation including status, stats, and
solution for science gateway development based on open
accessibility.
standards including JSR 168 and 268, OpenLDAP, SAGA,
and PKSC-11 [35]. The framework is still relatively new at the
Since its initial release in November 2011, the Foundation
time of this writing and as such, could not be included in the
API has supported over 250 unique projects representing 10k
evaluation process leading up to the development of the
scientists worldwide. Users burned nearly 9M SUs running
solutions described in this paper. Based on early successes,
over 10k jobs, leveraging 200 application codes installed on
CHAIN seems like an exciting project to watch going forward.
HPC systems at PSC, SDSC, and TACC. Version 2, due out
The target audience and advertised use case, however, are
prior to the publication of this paper, will add the following
more in line with traditional portal development than
services as well as expanded support for system registration,
lightweight gateways creation. The gUSE project provides a
federated identity management, additional execution
mature web service framework for running workflows, storing
platforms, and a more mature callback system.
data, and registering applications [36]. Further, it has existing
integration with the WS-PGRADE portal to provide an out-of-
• Systems: discovery and register storage,
the-box front end based on Liferay. As with CHAIN, the
authentication, and execution systems for use
PGRADE and gUSE project timelines ran parallel to that of
throughout the API.
the work in this paper. Futhermore, the approach taken by
• Transfer: move data from anywhere to anywhere
gUSE to provide a SOAP-based service stack runs counter to
using multiple protocols.
the desire of current web developers to interact with REST
services in an asynchronous manner. • Metadata: create, search, and infer metadata about
any resource (file, job, person, system, etc.) within
In response to the dearth of platform APIs available for the API.
general science the iPlant Collaborative created the
Foundation API [37]. The Foundation API is a RESTful By hiding all the heavy lifting of accessing systems,
Science-as-a-Service platform for building modern moving data, running simulations, and establishing
applications. It includes services that allow consumers to relationships between people, data, and devices, consumers
securely conduct science, manage data, and share and curate are freed up to focus on their science and developers are able
their work. Foundation exists as a hosted, multi-tenant cloud to focus on innovation at the application layer rather than
service that is freely available to the open science community. infrastructure at the system level.
Version 1 of Foundation supports the following services. V. YOUR NEXT SCIENCE GATEWAY
• Apps: Allows users to register and discover scientific Turning back to Andre’s GeneSeqr form, this tool is as
codes that can be run via the Jobs service. There are basic an example of a science gateway as one can find, but it
currently over 160 scientific codes both public and gets the job done. A scientist with remedial programming
private that can be run across multiple high capabilities can stand up a static web page on their personal
performance compute systems. computer, a public web server, or on a CDN such as their
public Dropbox folder, Amazon S3, or even a free Yahoo
• Auth: token-based authentication service. Issues
Sitebuilder page. When technology is that easy to adopt and
limited use tokens that can be restricted to a
reuse, the possibility for it to reach a broad audience increases
timeframe and number of uses and revoked when
dramatically. The question then becomes, how can we build
needed.
tools to accomplish tasks requiring a bit more complexity and
interaction and yet make them as simple to adopt and reuse as
Andre’s GeneSeqr form?
In recent years, a variety of toolkits and frameworks for We selected Backbone as platform for multiple reasons.
developing modern web applications have emerged that aid in Backbone adheres closely to our build-for-today design
the development of lightweight, responsive, standards-driven, philosophy. It is specifically designed for developing rich, yet
front-end components. These projects are open source, have lightweight client-side applications that utilize a RESTful API
very large user communities, and are supported by real backend. Backbone applications follow the Model-View-
companies such as Twitter (Hogan.js, Bootstrap) [38][39], Controller (MVC) design pattern making for code that is easy
DocumentCloud (Backbone.js, Underscore.js) [40][41], and to develop, maintain, and extend. As a JavaScript application
Google (Yeoman.io) [42]. Furthermore, as HTML5 has come framework, it can be easily integrated into other environments
into its own and the development of single-page applications and web platforms such as Liferay or Drupal. Finally,
has become commonplace in the commercial web, it makes Backbone is a widely used and popular framework with an
sense to begin using these technologies in science gateways. active user community and multiple examples of large-scale
applications built on top of it. Two examples of large-scale
With the goal of building tools that are simple to adopt Backbone users are the Khan Academy [43] and Coursera
and reuse in mind, we have developed a toolkit using these [44], both providers of massively open online courses
frameworks with the intention of doing for science gateways (MOOC). One can imagine the benefits of having a
what jQuery did for JavaScript and Web 2.0. We leverage the computational science course with labs and homework that
iPlant Foundation API as a backend, and provide plugins for included hands-on access to a computational environment
Backbone.js that allow a Backbone.js application to easily use where students can gain experience using actual large-scale,
the Foundation API without the developer needing detailed high-performance computational systems. Using the
knowledge of its inner workings. These plugins provide Backbone-Foundation plugins we have developed, these
implementations of the objects in the API as Backbone MOOC providers could easily integrate the Foundation API
Models and Collections that can be readily used to build into coursework offered through those sites.
science gateways. This allows the gateway developer to focus
more on gateway development and less on handling the web
service calls to the backing API.
TABLE 1. THE FULL LIST OF FOUNDATION API BACKBONE.JS PLUGINS
AND THE FUNCTIONALITY THEY PROVIDE.
Plugin Name Functionality Provided
backbone-foundation Core support for using Foundation API
backbone-foundation-apps Application discovery and registration
backbone-foundation-data Data transformation and staging
backbone-foundation-io Data management and movement
backbone-foundation-jobs Job submission and monitoring
backbone-foundation-profile Identity management
backbone-foundation-systems Resource discovery and monitoring
backbone-foundation-post-it Pre-authenticated URL shortening
The Backbone-Foundation library is broken into separate
plugins that can be included in an ad hoc manner based on the
needs of the application. At the core is the main backbone-
foundation.js file, which provides functionality for basic
interaction with the Foundation API. By providing extensions
of the default Backbone Model and Collection objects, the
Foundation API can be used transparently through the Figure 1. A standalone boilerplate gateway built using Backbone.js and
the Backbone-Foundation plugins. This application leverages the iPlant
standard Backbone API. Also in this plugin is an Foundation API to provide authentication, data management, application
implementation of the Foundation Auth API and Model discovery, and job submission with no backend other than the
objects for authenticating and obtaining API tokens for Foundation API.
authenticated use of the API. Finally, we include an Events
object that can be used to manage API-aware events across the We have also developed a complete Backbone application
application. (Figure 1) as a boilerplate science gateway using the
Backbone-Foundation library. The Backbone-Foundation
Support for the remaining Foundation services is provided library and Foundation API are white-label components that
by the additional Backbone plugins listed in Table 1. Each can be readily and easily used to develop your own science
plugin depends on the Backbone-Foundation core library to gateway. This application is built using Backbone for the
provide the API integration. The only other dependency of the application framework, Twitter Bootstrap for the front-end
Backbone-Foundation library is Backbone.js itself. components and HTML structure, and has no backend other
than the Foundation API and a web server to host the static
assets (which could also be hosted out of the Foundation API).
Figure 2. Embedding gateway widgets as a page in privately hosted CMS. From left to right: Wordpress, Drupal, Joomla, and Liferay sites.
The development of this boilerplate application took a single development and deployment of new features in a results-
developer less than a month to complete and includes driven fashion no matter how established an existing portal
authentication, data management, application discovery, and may be.
job submission.
Consider the example of the Liferay enterprise platform.
Lastly, we have developed a collection of embeddable Just as with the CMS platforms mentioned above, one can
“widgets” that provide discrete slices of functionality that can drop in a Foundation application using only HTML and
be used immediately to add advanced capabilities to any web JavaScript, and utilizing the Liferay Web Content Display
page or existing portal or gateway with no more effort than portlet as shown in Figure 2. Or, if something more robust is
adding a Twitter “Tweet this” or Facebook “Like” button. needed, the application can be wrapped in a portlet and
deployed it in the same way one would deploy any custom
These widgets are also built on top of the Backbone- portlet.
Foundation plugin library. To include a widget in a page, the
page author only needs to add a reference to the widgets script Whether this functionality is packaged as content (HTML
and a single div tag with the widget configuration contained in and JavaScript) or as a plugin, module, extension, or JSR 268
data attributes on the tag. The widgets script acts as a scout portlet for a specific platform, migrating the functionality from
script to discover the widget div, determine the desired widget, a lean prototype gateway using these tools, to an enterprise
and then inject the appropriate widget into the page. solution with all the bells and whistles is a trivial process.
Deploying features built entirely on the front end is not a
Foundation widgets can be easily used in any HTML page deliverable that consumes months of time and effort. On the
and many CMS platforms such as Wordpress, Drupal, or contrary, it is more akin to migrating static content from one
Joomla (Figure 2). And because they leverage the Foundation site to another.
API backend and don’t require local server configuration to
use, they can be used even on cloud-hosted sites such as Finally, as mentioned above, forward integration isn’t
Wordpress.com. limited to wrapping bits of functionality as pages. It is possible
to embed custom widgets to provide one-off functionality such
The widgets available at the time of this writing include a as activity streams, share buttons, data drop boxes, submission
drag-and-drop file uploader, an application discovery widget, forms, and directory trees just to name a few.
and a job execution widget. The uploader widget gives a drag-
and-drop upload functionality using the HTML5 FileAPIs The process of embedding a widget is the same as that of
allowing users to drag files from their desktop into the web adding a page. However, for easier adoption, an AJAX driven
browser in order to upload to the iPlant Data Store. The widget generator is provided on the Foundation API
application discovery widget allows embedding up-to-date developer’s website to help users create widgets based on their
lists of available iPlant applications into any page essentially unique constraints such as styling, default values, and
providing an application catalog for browsing and searching restricted permissions.
applications. The job execution widget allows the embedding
of an application-specific job submission form in any page. VII. CONCLUSION
Science gateway development has historically been an
VI. GETTING FROM TODAY TO TOMORROW
enterprise effort. In recent years, the introduction of
The discussion on building for today has been targeted at lightweight web technologies and REST APIs have changed
researchers developing new gateways up to this point. the way modern applications are built. By leveraging the
Previous sections have demonstrated how one can bootstrap technologies of today and decoupling complex infrastructure
an idea into a functional science gateway with a relatively from gateway front ends, developers can respond to change
short ramp-up using existing APIs and services like the faster, innovate more quickly, prototype more easily, and
Foundation API. However, these same development principles drastically reduce their time to production. This paper presents
can benefit existing gateways and portals, enabling the rapid a set of reusable, white labeled, front end components written
entirely in HTML, JavaScript, and CSS that leverage the [15] 2005. Edublogs – education blogs for teachers, students and schools.
http://edublogs.com/.
Foundation API and enable just such a transformation. By
[16] "nsf.gov - National Science Foundation (NSF) News - XSEDE Project
utilizing the backbone-foundation plugins as fully functional, ..." 2011. 29 Mar. 2013.
interchangeable components, both new and existing gateways http://www.nsf.gov/news/news_summ.jsp?cntn_id=121181.
can shift their attention from tedious integration to rapid [17] 2002. Liferay.com: Enterprise open source portal and collaboration
innovation that can impact researchers today rather than software. http://www.liferay.com/.
tomorrow. Both the gateway components and the Foundation [18] Dahan, Maytal, Eric Roberts, and Jay Boisseau. 2007. TeraGrid User
API are freely available for use today at Portal v1. 0: Architecture, Design, and Technologies. International
Workshop on Grid Computing Environments. November 28.
https://foundation.iplantcollaborative.org.
[19] Martin, Robert Cecil. 2003. Agile software development: principles,
patterns, and practices. Prentice Hall PTR, September 1.
ACKNOWLEDGMENT
[20] "Instagram." 2009. 29 Mar. 2013. http://instagram.com/.
The iPlant Collaborative is funded by a grant from the [21] "Pinterest." 2009. 29 Mar. 2013 .http://pinterest.com/.
National Science Foundation Plant Cyberinfrastructure [22] "GivePulse | Enabling Everyone to Volunteer." 2012. 22 Mar. 2013.
Program (#DBI-0735191). This work was also partially https://www.givepulse.com/.
supported by a grant from the National Science Foundation [23] "World Wide Web Consortium (W3C)." 29 Mar. 2013.
Cybersecurity Program (#1127210). http://www.w3.org/.
[24] 2006. Adobe - Flash Player.
REFERENCES http://www.adobe.com/software/flash/about/.
[25] "CSS Transforms." 2012. 29 Mar. 2013. http://www.w3.org/TR/css3-
transforms/.
[1] Miller, M.A., Pfeiffer, W., and Schwartz, T. (2010) "Creating the
CIPRES Science Gateway for inference of large phylogenetic trees" in [26] "CSS Transitions." 2009. 29 Mar. 2013. http://www.w3.org/TR/css3-
Proceedings of the Gateway Computing Environments Workshop transitions/.
(GCE), 14 Nov. 2010, New Orleans, LA pp 1 - 8. [27] Schlueter, Shannon D, Qunfeng Dong, and Volker Brendel.
[2] Dooley, Rion, Kent Milfeld, Chona Guiang, Sudhakar Pamidighantam, "GeneSeqer@ PlantGDB: Gene structure prediction in plant genomes."
and Gabrielle Allen. 2006. From proposal to production: Lessons Nucleic Acids Research 31.13 (2003): 3597-3600.
learned developing the computational chemistry grid cyberinfrastructure. [28] "jQuery." 2006. 29 Mar. 2013. http://jquery.com/.
Journal of Grid Computing 4, no. 2: 195-208. [29] "ProgrammableWeb - Mashups, APIs, and the Web as Platform." 2005.
[3] Demeler, BORRIES. 2005. UltraScan: a comprehensive data analysis 29 Mar. 2013. http://www.programmableweb.com/.
software package for analytical ultracentrifugation experiments. Modern [30] 2012. 8,000 APIs: Rise of the Enterprise - ProgrammableWeb.com.
analytical ultracentrifugation: Techniques and methods: 210-229. http://blog.programmableweb.com/2012/11/26/8000-apis-rise-of-the-
[4] Goecks, Jeremy, Anton Nekrutenko, James Taylor, and T Galaxy Team. enterprise/.
2010. Galaxy: a comprehensive approach for supporting accessible, [31] "Backend as a service - Wikipedia, the free encyclopedia." 2012. 29
reproducible, and transparent computational research in the life sciences. Mar. 2013. http://en.wikipedia.org/wiki/Backend_as_a_service.
Genome Biol 11, no. 8: R86.
[32] "Twilio Cloud Communications - APIs for Voice, VoIP and Text ..."
[5] Klimeck, Gerhard, Michael McLennan, Sean P Brophy, George B 2005. 29 Mar. 2013. http://www.twilio.com/.
Adams, and Mark S Lundstrom. 2008. nanohub. org: Advancing
[33] 2007. Soaplab2. http://soaplab.sf.net/.
education and research in nanotechnology. Computing in Science &
Engineering 10, no. 5: 17-23. [34] Cholia, Shreyas, David Skinner, and Joshua Boverhof. 2010. NEWT: A
RESTful service for building High Performance Computing web
[6] Norris, Pippa. 2003. Digital divide: Civic engagement, information
applications. Gateway Computing Environments Workshop (GCE),
poverty, and the Internet worldwide. Vol. 40. Cambridge: Cambridge
2010. IEEE, November 14.
University Press, September.
[35] CHAIN, (2010) Co-ordination & Harmonisation of Advanced e-
[7] 2012. Science and Engineering Indicators 2012.
Infrastructures EU FP7 project http://www.chain-project.eu) project
http://www.nsf.gov/statistics/seind12/start.htm.
[36] Peter Kacsuk, Zoltan Farkas, Miklos Kozlovszky, Gabor Hermann,
[8] Bureau of Labor Statistics, U.S. Department of Labor, Occupational
Akos Balasko, Krisztian Karoczkai and Istvan Marton:
Outlook Handbook, 2012-13 Edition, Medical Scientists,
WS-PGRADE/gUSE Generic DCI Gateway Framework for a Large
on the Internet at http://www.bls.gov/ooh/life-physical-and-social-
Variety of User Communities
science/medical-scientists.htm (visited May 10, 2013).
Journal of Grid Computing, Vol. 9, No. 4, pp 479-499, 2012.
[9] Fawcett, Henry. Manual of political economy. Macmillan, 1888.
[37] Dooley, Rion, Matthew Vaughn, Dan Stanzione, Steve Terry, and Edwin
[10] Wilkins-Diehr, Nancy, Dennis Gannon, Gerhard Klimeck, Scott Oster, Skidmore. Software-as-a-Service: The iPlant Foundation API.
and Sudhakar Pamidighantam. 2008. TeraGrid science gateways and
[38] 2013. Hogan.js - Twitter on GitHub. http://twitter.github.io/hogan.js/.
their impact on science. Computer 41, no. 11: 32-41.
[39] 2011. Bootstrap. http://twitter.github.com/bootstrap/.
[11] Lawrence, Katherine A., and Nancy Wilkins-Diehr. "Roadmaps, not
blueprints: paving the way to science gateway success." In Proceedings [40] 2011. Backbone.js. http://backbonejs.org/.
of the 1st Conference of the Extreme Science and Engineering [41] 2008. Underscore.js. http://underscorejs.org/.
Discovery Environment: Bridging from the eXtreme to the campus and [42] 2012. Yeoman - Modern workflows for modern webapps.
beyond, p. 40. ACM, 2012. http://yeoman.io/.
[12] Wilkins-Diehr, Nancy, and Katherine A. Lawrence. "Opening science [43] 2009. Khan Academy - Wikipedia, the free encyclopedia.
gateways to future success: The challenges of gateway sustainability." In http://en.wikipedia.org/wiki/Khan_Academy.
Gateway Computing Environments Workshop (GCE), 2010, pp. 1-10.
IEEE, 2010. [44] 2012. Coursera. https://www.coursera.org/.
[13] "XSEDE | Overview." 2011. 29 Mar. 2013.
https://www.xsede.org/gateways.
[14] 2005. Joomla! The CMS Trusted By Millions for their Websites.
http://www.joomla.org/.