Skip to content.
Sections
Personal tools
You are here: Home » Features » Developing an SOA at SUNY; Lessons learned

Developing an SOA at SUNY; Lessons learned

Christina Smart
Last modified 19 Mar, 2007
Published 23 Jan, 2007
We spoke to Patrick Masson former Director of Technology for the SUNY Learning Network about what happened when they tried to implement a Service Oriented Architecture at the State University of New York, an institution with 64 campuses, 30,000 faculty and 414,000 students. Patrick reflects on the lessons he’s learned and how he plans to implement Service Oriented Architectures in his new role as CIO at SUNY Delhi College of Technology. It’s quite a long interview but will be of great interest to anyone thinking about implementing SOA in their institution.

Introduction

The technical cornerstones of the e-Learning Programme are open standards, web services and service oriented approaches [1]. Through the toolkits, demonstrators and reference model projects much has been learned in the last two years about using these technologies and there have been some significant successes like the XCRI course description schema [2]. Much of the work so far could be said to have been blue skies and experimental, but many are predicting shifting to SOAs will have a big impact on business and education:

“Service-oriented architectures (SOAs) are going to change everything. From software architecture to development to deployment, over the next few years SOAs will have a bigger impact than most modern paradigm shifts. It'll be bigger than the shift to the Web..” [3]

So what is required to make these technologies mainstream? At this point in time examples of early adopter institutions are very useful, in particular those that are prepared to implement SOAs and share their experiences with the community [4].

One such institution is SUNY (State University of New York) which two years ago put together a team to review its provision through the SUNY Learning Network (SLN) [5], [6]. What they came up with was a radical new approach based on the JISC/DEST e-Framework for Education and Research [7]. We spoke to Patrick Masson former Technical Director of the SLN about the project and why the SOA approach his team was advocating ran into difficulties. It will not surprise many that the most significant barriers were organisational ones. Patrick also talks about how he is building on his experiences at SUNY in his new role as Chief Information Officer (CIO) at SUNY Delhi (Del-high) [8].

Interview

CS: Can you tell us about SUNY and the SLN and what the project was trying to achieve?

PM: SUNY is a large and diverse organisation, it has 64 campuses which range from community colleges to research universities. Each campus has a different mission and a different e-learning strategy and the campuses are pretty autonomous. SUNY Central Administration provides assistance in the development of technologies and initiatives with system-wide or broad appeal, and one of these was the SUNY Learning Network (SLN) which began in 1994 at Empire State College for their online distance learning using Lotus Notes/Domino [9]. SLN was a grassroots programme which became successful and grew to include about 40 campuses delivering entirely asynchronous distance learning courses. At the same time many of the web based learning management systems emerged and allowed campuses to exploit blended approaches, where face to face courses could be enhanced with use of the web.

What were the issues with using Lotus Domino?

There were a few issues that emerged and some of these weren’t made very clear to the SUNY campuses.

The SLN Lotus Domino system was delivering online learning to 40 different campuses all of which had their own student information systems (SIS) as well. Whereas SLN was running Lotus Domino centrally by system administration, the SISs were being run locally by each different campus. This made things difficult for campuses to administer their on campus and online courses– there is no central SUNY student identity for example – so students had to go through a double registration, one for on campus courses through the local SIS and one for SLN courses which obviously caused problems for campuses: double enrolments, enrolments in the wrong course section, etc

It also meant that single sign-on and SIS integration was an issue. Many campuses had their SIS connected to their local LMS (WebCT, Blackboard, etc.) so that campus courses were populated automatically in the LMS with student data. Campus administrators wondered why they couldn’t have the same functionality with SLN.

Another issue for SUNY Administration was the customisation of Lotus Notes and Domino. SLN had some 100,000 enrolments managed through Lotus Notes and Domino and had customised the system to such an extent that only a few people knew how it worked. IBM couldn’t support SLN's customizations. This meant that central administration, all of SLN as well as the 40 campuses, was dangerously dependant on a couple of individuals.

But there were positives to running a central service as well, for example cross registration. If a student enrolled with SLN they could see the entire SUNY online catalogue of courses available from all campuses – not just their own.

So SUNY Administration decided to evolve SLN away from using Lotus Notes/Domino and put together a team to develop the next generation of SLN. The team included myself from UCLA, Ken Udas from the Open Polytechnic of New Zealand and now Executive Director of Penn State's World Campus, Michael Feldstein now product manager for Oracle's Academic Enterprise Initiative, and Bernie Durfee who had worked on uPortal. The idea was to develop something that retained all the advantages of the SLN while eliminating the problems. The team developed the SLN 2.0 strategy which would remain a centralised system, but would offer a diverse set of tools for the different campuses [10].

So it was quite teaching and learning focused?

Very much so. Michael Feldstein made the argument of the long tail, that educational applications are best developed through collaboration [11]. For example at Suffolk County Community College they have a telescope that can be controlled over the web and students can go out and log what they see in the night sky. The system we developed should be able to allow that tool to be available to all the SUNY campuses over the web. We believed that each course should be able to select the tools that they need.

So we decided to move away from tool development – because that was already happening in lots of different places, the SLN 2.0 strategy was about providing a framework for those tools to plug in to. This was at a time when the commercial LMSs were offering more services, like card swiping and e-commerce which makes them more attractive to senior executives as the complete solution, yet provides little in the area of teaching and learning.

What were the advantages of the framework approach over Sakai?

I think what happened with the Sakai project was that it needed to build an LMS to compete with Blackboard which meant it needed to lock down everything and produce a fully functioning system so it could be adopted [12]. But there were examples of where universities were using some of the Sakai tools, e.g. the Sakai Gradebook [13] with other tools and integrating them via uPortal [14].

So our blueprint for SLN 2.0 was to break out Sakai's tools, combine them with ‘best in class’ tools then present them all through uPortal. DeAnza Community College had done this, replacing Sakai's discussion forum with Jforum and Alfresco. We were also talking to James Dalziel about LAMS because the presentation and learning design approach of LAMS fit the culture of teaching and learning of SLN [15].

We also wanted to take on a versionless approach with our development. Because SLN was a remote service developed and managed centrally our end-users never needed to install upgrades or migrate to new versions. This would allow us to introduce new features incrementally with small iterations that extended functionality. Every time a user logged in they might see a small bit of new functionality. To highlight this versionless approach I often asked, “What version of Yahoo are you running?” This meant that there would be lower implementation costs – because there would be a gradual change – not a big flipping of a switch and roll out of a brand new system, with all the issues that that entails. So we were into agile software development and incremental approaches that are behind current web developments.

The problem was that at the same time the SUNY administration was looking to make a big statement to ease the anxiety of campuses who had heard for many years, “Lotus Notes/Domino is going away”. This strategy (a components based framework) and our approach (Iterative and Incremental Design) wasn't an easy answer, like “We're migrating to Angel”. I think our idea was to ease anxiety by causing as little trauma through change as possible, the Administration's idea was to ease anxiety by picking something everyone knew. So the administration didn’t like the SLN 2.0 strategy because it couldn't be easily communicated as an “out-of-the-box” solution, you had to understand then educate the users about new concepts like SOA, loose-coupling, data hubs, service buses and message brokers. We ran into the “Build vs. Buy” argument and fears over open source. We lost our support from senior management.

For the community the SLN 2.0 blueprint wasn’t the norm. They were used to systems like Blackboard and Web CT so it didn’t fit their model of how things work. IT managers in campuses wanted to offload all their administration issues onto central administration while at the same time wanting something totally tailored to their needs.

Service Oriented Architectures came into it because we were going to treat every campus student record system as a legacy system, and route the student and course data which we could use to an abstract data object. We were also influenced by the Sun e-Learning framework [17], the JISC/DEST e-Framework, SCT Luminis Data Integration Suite [18], and IBM Enterprise Service Bus [19].

So this was the only way we thought we could deal with all the needs of the 64 campuses, but we were very unsuccessful at getting the message across. And to be honest with you the marketing people from the big providers are much better at selling this stuff than we are. The problem is that SOA represents a complete change in how traditional systems are developed, produced and rolled out.

So how do you go about telling people about the benefits of service oriented architectures?

That’s when you need to present use cases. One good example is dates and calendars. Think about all the times a calendar is used. As a student you have one calendar for when assignments and assessments are due in your course, and one for your departmental dates. Then you’ve got your personal email calendar with your personal events in as well as a campus calendar with term dates in. That’s four calendars right there that you are expecting students to check and use. Wouldn’t it be great if all that information was aggregated into one calendar? The only way is through a SOA. That’s one of the first things we were going to do with the SLN 2.0.

There was some discussion about whether we were going to integrate this data at the portal presentation layer or whether we were going to have true interoperability and data integration and activity so all the data could display in all the different systems.

So there’s a fundamental difference between monolithic enterprise approach and a service oriented architecture approach where calendar is just an available service and it doesn’t matter which one we use as long as we have open and accepted standards. And if there is a standard for date format then that information can be pulled into other applications like grade books.

And we did it! We took the Bedework Calendar and Samigo Test Engine [20], CAL’s Sakai Gradebook and legacy SLN tools running in Notes/Domino and not only integrated them so that they all appeared as a unified LMS, but we also provided interoperability where data generated in one tool was shared with other tools . This was done through the development of what we called the “abstract data object” where all four tools stored and accessed data.

Do all the standards that are needed exist yet?

No. IMS is trying to develop more but all the standards needed aren’t there yet [21].

In SLN 2.0 we had decided to start by developing a message broker and a service bus (registry) but we also took on the crusade of developing more standards. We started a conversation in the US about the Learning Management Operating System and we wanted to develop use cases about how people want to push information from tool to tool. The key questions were - how should messaging happen? and what sort of tool attributes are needed?

So where is the concept of the Learning Management Operating System now?

At the recent Educause meeting [22] in the US we organized a meeting of several campuses (MIT, California, Penn State, SUNY, Mary Washington), companies (Oracle, SUN, Desire2Learn, Moodlerooms, Unicon) and even organizations (Sakai, Mellon Foundation) to discuss the future of what has become known as a Learning Management Operating System (LMOS).

Everyone agrees that it needs to happen, they understand it because they run into it everyday. They are all integrating systems. But there’s not a lot of pay off for doing it – because they are all working with campus environments where they have limited budgets and limited staff and production environments.

I wonder why Sakai isn’t devoting its time and energy to producing a framework which at the end of the day many people could use, what do we gain from getting another Moodle?

One of the JISC projects suggested to me recently that building a SOA from scratch might be too hard for institutions. What’s your view?

It’s a huge undertaking to persuade CIOs and campus presidents to abandon a production and deployment model - the Enterprise Resource Planning (ERP) way of doing things. People don’t have the technical background and understanding. The ERP allows people to showcase systems and say: “wasn’t our money well spent?”, whereas an incremental and bottom up approach just doesn’t allow management with those showcasing opportunities.

There have been lots of white papers, for example from IBM, but it’s scary for institutions to take on this new approach. It involves agile project management and an incremental problem solving philosophy. Everyone has different hot topics that need addressing, this is a new model for solving people’s issues – rather than buying a new system. I want to find small success stories and use them as case studies. Bedework Calendar is one example. We don’t have to solve all the problems right now. Everyone is stymied by the big picture – I’m saying let’s start small and solve this one small problem, like calendaring, with this new approach.

Is SOA inevitable? Will it just happen in education and business?

I think it’s already happening and it might be forced on academics. Look at Oracle and the Academic Enterprise Initiatives [23], IBM and SCT Sungard. At some point the number of tools that are available will get to a threshold point.

Look at my college, a small rural campus of 2800 students but already computing expectations are greater than we can provide. I wonder whether we (institutions) will just be by-passed – people are already creating their own study groups in Facebook [24] and forwarding their campus email to their own email accounts. How can traditional or enterprise systems keep up with that? And as soon as systems have been adopted and rolled out there’s a new medium and operating standard available that everyone is using for sharing information and contacting each other.

So institutions will find themselves being by-passed by students using web applications?

Students and faculty. Look at the wiki and the blog. Look at Moodle’s success [25] – how many of those deployments started life as an official top down implementation versus how many started life as a rogue implementation by an individual academic in their department? That’s the environment we’re living in and I think an agile development approach and a SOA approach fits that real world academic computing environment.

But there are problems with this bottom up approach, institutional buy-in and funding are an issues for development teams. Evidence-based development is different than administrative initiatives. If developers want to devote time and energy to an emergent project, to document it and go and talk to other people about it, whatever, and it is not part of the ERP roadmap, they will probably receive little institutional support. How do you invest in these grassroots developments?

I guess we’re lucky in the UK that the JISC is funding these types of developments through the e-Learning Programme.

What JISC and CETIS are doing is exactly what should be happening, by funding ten £40,000 projects instead of one £400,000 project. Mellon foundation has just done a similar thing – the Bedework Calendar has just received some funding from Mellon [26]. And look what they’ve been able to do from a focused needs based calendar development, but with the right philosophy behind it.

Do you want to talk about what you are doing now at SUNY Delhi?

I’m Chief Information Officer at Delhi, a small rural college which is one of the SUNY campuses. The job was interesting to me because of the college's traditional approach with some legacy systems and is looking for new ways of doing things. I thought it would be interesting to do the sorts of things I’ve been talking about, taking a grassroots evidence-based agile approach to development. The administration is very open to this approach so it’s a great opportunity. They’d like to provide the sorts of opportunities that are available at other institutions. So I'm introducing new practices based on Agile Project Management methods to undertake their traditional approach to needs analysis to identify new projects, project planning, development, etc.

But there’s a lot of education that needs to take place, there are the technical aspects of SOA – but there are also the cultural aspects of SOA. I’m trying to start with the cultural aspects this time, and to create a culture where this agile SOA approach is the norm. The question is can we get planning and management to follow the agile SOA type approach, instead of the three year enterprise planning type budget approach where decisions are made through a committee? And I’m including other services as well, like helpdesk services.

So based on your experiences what would your advice to institutions be?

Well I’m not sure institutions are thinking about these things – I think people in institutions are thinking about these things. The developers who have to work every day in this are thinking this way. I still hear lots of colleagues at Educause meetings happily extolling the virtues of their latest implementation of the latest monolithic system. At the high level people still don’t get it. How many senior managers in charge of technology on campus have a technical background? They might be procurement officers that are put in place to control the costs of technology and find single solution systems for the lowest price. Or maybe they are faculty and they are now responsible for e-learning and IT – how many information technologists are now Chief Information Officers? Not many. And where does the CIO fit in the overall management of a campus? Fundamentally there is a conflict between where we (the developers) are and where the decision makers are.

More information on Patrick’s work at SUNY Delhi can be found on his blog [27].

References

[1] The e-Learning Programme

[2] Exchanging Course Related Information (XCRI) project

[3] The SOA Sea Change, Alan Zeichick

[4] Large Scale deployment of web services at London Met

[5] State University of New York

[6] SUNY Learning Network

[7] JISC/DEST e-Framework for Education and Research

[8] SUNY Delhi College of Technology

[9] IBM Lotus software

[10] SLN 2.0 strategy (pdf document)

[11] Michael Feldsteins blog 30 Jan 2006 explaining LMOS and the SLN teams approach

[12] Sakai Project

[13] Sakai Gradebook Tool

[14] UPortal

[15] LAMS, Learning Activity Management System

[16] Bedework Calendar

[17] Sun Microsystems e-Learning Framework

[18] SCT Luminis Data Integration Suite

[19] IBM Enterprise Service Bus

[20] Samigo Test Engine

[21] IMS Global Learning Consortium

[22] Educause

[23] Oracle Academic Enterprise Initiatives

[24] Facebook

[25] Moodle

[26] Andrew Mellon Foundation

[27] Patrick Massons blog

 

Supported by JISC Supported by CETIS
Powered by Plone