Alternative Architectural Concept 2 - Federated
Integration
By Stephen Lahanas, E-Learning
Leaders
November 12, 2002
This is the second in a series of articles discussing possible variations to
the Learning Technology Systems Architecture (LTSA) model delineated by the IEEE
Learning Technology Standards Committee (LTSC) working group. In a previous
article I covered one of the core structural issues / content object taxonomy.
With the taxonomy I implied some deeper philosophical predispositions which I
feel are vital to the success of any potential learning architecture, they
include the following assumptions:
- Learning (or e-learning) is a process (which can be thought of as a
superset with a nearly infinite number of subsets contained within)
- Data / information without "personal transformation" is not and cannot be
knowledge. Personal knowledge therefore represents the ultimate end-product of
any learning process
- Learning process variation should not be restricted, any successful
architecture must empower the process/es, not narrow them.
- Management of knowledge is in fact a learning process.
- Management of data without human (or AI ?) learning of the data is a
mechanical function and adds little value. A library is an example of how
knowledge management should work – the library is a physical repository of
filtered / contextualized data ready for personal processing. The process
though that the library serves is learning.
- Tightly coupled integration does not work well and costs as much as 10x or
more as much as a loosely coupled paradigm.
- The Internet provides us the infrastructure for a learning architecture or
framework, but does not represent the model for it necessarily. Learning
architectures are an evolution perhaps of the Internet model to its next
logical state. E-learning represents the best framework approach for advancing
the Internet model.
Federated or Distributed ?
The last point
referred to the Internet model as one that may be improved on rather than
followed. In previous discussions on Leaders we’ve tried to define that model a
bit, and the term 'distributed' has been used to try to quantify it. I think
though that "distributed" perhaps doesn’t capture the true spirit of the
Internet model – I’d think of it as Federated rather than distributed. The word
‘distributed’ implies that one system will be spread out to different locations.
Federated supports the sense that we’re dealing with a host of systems, (which
in truth are not really components if they can stand alone). These systems form
a "cooperative" community which can be morphed into many manifestations simply
by inclusion or exclusion of end-user configuration / access to
them.
"Federated" then implies a higher level of abstraction and autonomy
in the various systems which could be aggregated into a learning solution
architecture. This in fact is an excellent evolutionary metaphor for the
Internet. Having worked with enterprise integration, component development and
web development, I can state with confidence that understanding the difference
between these concepts is vital to the success of any effort in this regard.
Proper logical conceptualization at this stage of what we think a "global"
learning architecture should be, will make all the difference as to whether or
not we can pull it off anytime soon.
Systems or Components
?
Sometimes this may sound like a debate in semantics, but it’s
not.
A component is generally considered a piece of a system, whereas a
system contains components and operates in a stand-alone mode. A component
however has no guarantee of operating in a stand-alone fashion and is usually
dependent on something else to add value. Any time we’re dealing with a global
architecture, (which is very different from what LSTA had envisioned, but close
to the types of distributed repositories that Stephen Downes has been
describing), we must consider that component connectivity across firewalls and
distances is simply not efficient.
In manufacturing or electronics a
component is generally considered to be a semi-autonomous or completely
dependent sub-assembly of a closed system. Now let's make an analogy with
biological systems that will help further clarify the matter. Within a body
there are many components (organs) that are required in order for the system
(body) to function - those components are dependent, sometime redundant and
highly specialized but not autonomous in that they are directed centrally
(neuro-motor control). The components then are also sub-assemblies and cannot
standalone, yet they are distributed geographically within the closed body
system. Each closed body system however is autonomous and participates in a much
larger eco-system of other autonomous systems which together comprise various
logical frameworks (such as business a government etc). This is the difference
between federated and distributed - autonomous systems within a system of
systems versus dependent sub-assemblies of a single system.
What we see
in any viable global architecture is the need to segregate application
functionality and provide efficient data flows between them. The segregated
application functionality then necessarily becomes a stand-alone application or
service – more or less tantamount to a system. So we have a global network
comprised of independent networks with a global learning architecture comprised
of independent yet federated systems.
Loose Integration
By
nature then, autonomous systems cannot and should not be tightly integrated in
that unnecessary dependencies will arise that increase complexity and reduce
efficiencies. The problem is that most people in the information technology
arena have been focused on delivering tight integrations for the past decade or
more. There is a term for this - Enterprise Architecture Integration (EAI). An
enterprise is in itself a mini-system of systems scenario but one limited to one
organizational domain (often times there are many networks + web access as well
so it isn't a one or closed network scenario). The fact that there is a single
domain to integrate leads most to follow what may have worked reasonably well
for client server technology - that is a consolidation and tight coupling of the
application layer into interdependent components.
In the 1990s, the main
technology used to facilitate this type of integration was CORBA although many
other less rigid (more proprietary approaches were attempted). To describe all
of this outside the technological climate would be a bit misleading, there were
very good reasons for wanting to split applications / systems up in the '90s.
The first move to client / server technology went from mainframes to mid-tier
servers with Pentium 200MHz clock speeds and 100 MB hard drives. The hardware of
the early and mid-'90s simply did not support deployment of powerful
applications - they had to be spread out (a great motivation for early AI
efforts that led to MPP "Massively Parallel Processing" the technology behind
Teradata data warehousing for example).
What we found out though by the
end of the '90s was the complexity of your average IT department that built
interdependent systems across a weak mid-tier was consuming 1/2 or more of the
yearly budgets to maintain, many IT departments are still stuck there. The
methodologies and philosophies of that era are still with us in spirit anyway -
one gets the real illusion of control when engaging in a tight integration
effort and in fact that illusion becomes a necessity once embarked on that path
else the solution simply won't work. (but then you've spent the majority of your
dollars not on the requirement but on what was perceived to be a simple
integration - unfortunately that's how they're always perceived).
Moving
to expand on an Internet or federated model requires a paradigm shift in IT
thinking. The entire success of e-learning as a global solution depends on this
shift.
Paradigm Shift
The key to any shift is the creation
of a global understanding of the potential marketplace, Stephen Downes refers to
something similar, "the Learning Object Economy," which describes part of it
well but doesn't extend to the whole spectrum. On Leaders this Spring we took a
stab at trying to redefine the marketplace in such a manner that would allow
vendors and potential users alike to appreciate the potential that is now being
overlooked in an LMS-centric, myopic approach to e-learning. The first step is
to recognize the logical expansion of the term e-learning to encompass the
cross-section of opportunity where education and technology are already
co-located. This cross-section expands the current definition of the market by
approximately ten-fold.
Another crucial step is for current e-learning
luminaries to stop being so judgmental about what constitutes 'good' e-learning
or 'quality' e-learning as this has artificially elevated market costs and
planted unnecessary doubts and reservations in the minds of the very people who
are most interested in adopting the technology. E-learning, like the Internet is
about extension, expansion, access and new opportunity - naturally many feel
threatened by its potential, but it is an unwarranted fear that will quickly
disappear as adoption grows.
We won't dwell too long on marketplace issues
just now, but it is important to keep in mind that everything we do with the
technology / solutions will be guided by the assumptions we're working under.
The technical paradigm shift needed is to push vendors towards more specialized
development that will inter-operate at a federated level or serve a standalone
purpose. For example, content development, assessment activities, simulations,
labs, course, video-audio, content objects and games could all be accessed
singly and add value or be aggregated per rule and add value.
One of the
greatest follies in e-learning that I've witnessed (and I must admit, believed
in at one time) is the idea of the intelligent system deciding things for the
learner - this has been most often manifested through something called
'assessment based prescription.' The idea behind this is that based upon rules
built into a system after an assessment a learner will be directed to the exact
information they need automatically (they are of course many variations on this
theme). Here's the problem:
- Building it adds 10x the complexity to the project
- We make the rather arrogant assumption that whoever decided the rules for
all students in such scenarios will adequately support those students' needs.
I can guarantee you here today that it will not, the possible number of
variations and permutations necessary to suit every students' view into a
particular subject cannot be adequately predicted and moreover doesn't have to
be.
- So here's the revelation, letting the student have control actually
reduces complexity, increases effectiveness and will certainly improve
adoption. And this by the way, is the main interface process model for the
Internet, so we have some real proof of the
potential.
"Learner Centric"
A battle cry for
many years among educators - now we can actualize it in the most powerful manner
imaginable. What we are doing in sense is relieving a lot process duties from
integrated systems thereby allowing them to be loosely coupled and shifting that
burden where it always should have been, the learner who defines her or his own
processes. Now, this doesn't prevent us from providing help in those journeys,
but removing the need for massively centralized control removes the integration
burden that might otherwise occur - one that is strongly implied in the LTSA
architecture even at the enterprise level and would be more complex if extended
globally.
©Copyright 2002
Ed. Notes:
Many thanks to
Stephen Lahanas for allowing the feature to be republished here.
The
whole Alternative Architecture series can be found on the E-Learning Leaders Group
(requires a Yahoo username). The first
article deals mainly with object classification. The third
article outlines a Federated E-learning Architecture. More articles will
follow.
The IEEE-LTSC
Architecture and Reference Model Working Group has made the working draft 9 of
the LTSA (947 Kb, pdf) available on their website.