Wang GuiLing, Li YuShun, Yang ShengWen, Miao ChunYu, Xu Jun, Shi MeiLin
Computer Networking & CSCW Lab, Department of Computing Science & Technology, Tsinghua University
Keywords: Grid, CSCW, e-Learning, Web Services
Abstract: Large-scale cooperation support for learners becomes even more important, when e-Learning is
implemented in scalable, open, dynamic and heterogeneous environment. This paper presents how to
realize collaborative learning support in distributed learning environments based on grid technology. Our
approach fills the existing gap between current cooperative platform and complex, cross-organization
infrastructure. We propose the grid architecture for establishing collaborative platform for e-Learning,
where grid middleware and CSCW services are provided. A Learning Assessment Grid, abbreviated as
LAGrid, is built on top of these services and provides collaborative learning in large-scale cross-
organization environment.
It is still an unsolved problem in e-Learning research
that supporting collaborative learning in scalable,
open, dynamic and heterogeneous environment. The
scenario is a large-scale connected environment of
learning management systems, learning content
management systems and virtual classroom systems
of different organizations. Such an environment
may support various services such as widely
collaborative learning on top of kinds of
collaborative tools, collaborative usage of learning
resources, monitoring the learning activity of others,
feedback and notification of learning activity etc.
The target environment is scalable, by which we
mean that infrastructure and services can be
decoupled from their users and expanded on
Grid Computing has its origins in wide-area
distributed computing, and extends to a large-scale
“flexible, secure, coordinated resource sharing
among dynamic collections of individuals,
institutions, and resources” (Foster, 2001). Modern
grids also focus on scalability and openness.
Therefore Web Services technologies are introduced
and adopted. The implementation of OGSA (Foster,
2002) which is actually based on Web Services with
specific extensions has combined grid computing
and Web Services technology. Grids are typically
implemented as a form of middleware which
provides all grid-related services and can also use
the Internet as a communication infrastructure
(Atkinson, 2004).
Collaborative learning is taken as a fundamental
and essential learning activity (Cowie, 1988).
Application in which such activities are supported is
perceived as “collaborative platform for e-Learning”
in our view. CSCW technology is adopted to
support collaborative learning, and it is more and
more widely used such as in BSCW
( and ANTS
(Lopez, 2001) recently. However, up to now,
complex e-Learning applications which support
large-scale, cross-organization and collaborative
learning in an open, dynamic and heterogeneous
environment have rarely been considered in the
context of e-Learning or in CSCW, mainly due to
technical feasibility problems or high costs. Now
rapid evolution of grid computing gives us a chance
to integrate the technology into collaborative
platform for e-Learning.
As a result, in the “Grid-based Collaborative
Platform for e-Learning” project, we are working on
solutions for collaborative learning in grid
environment in the context of cooperation work
platform. We investigate both the impact of grid
infrastructure on CSCW theory and technology and
the requirement of CSCW application for grid
infrastructure. In particular, we investigate how to
make collaborative platform grid-enabled with grid
GuiLing W., YuShun L., ShengWen Y., ChunYu M., Jun X. and MeiLin S. (2005).
In Proceedings of the First International Conference on Web Information Systems and Technologies, pages 510-517
DOI: 10.5220/0001227705100517
techniques employed in the infrastructure, especially
methods and techniques developed for collaborative
learning. In the first phase of this project, we
developed “Distance Learning Assessment Grid”
(LAGrid) for learning assessment which is a critical
aspect in e-Learning.
In this paper we are to describe an approach of
building a grid-based collaborative platform in
distributed, dynamic and heterogeneous e-Learning
environment. In particular, we will employ the Web
Services and Services Oriented Architecture (SOA)
technologies to build grid middleware, combine
groupware technology with grid computing to
develop generic CSCW services, and study a layered
open architecture.
The organization of this paper is as follows:
First, we motivate our work by a scenario of open e-
Learning requirements. We propose the grid
architecture for domain application which contains
five layers: Infrastructure Layer, Basic Service
Oriented Architecture Layer, Grid Middleware
Layer, CSCW Layer and Domain Specific Services
& Application Layer. The detailed functions of each
layer are introduced in section 3. Section 4 shows
the implementation mechanisms of the main
components in Grid Middleware Layer, CSCW
Layer and Learning Assessment Layer. The paper
ends with conclusion and remarks on future work.
Grid-based Collaborative Platform for e-Learning
aims to develop a scalable collaborative
environment to meet the increasing requirement of
national education by integrating the learning
resources, software resources and educator resources
from above 40 autonomous organizations in China
Central Radio & TV University (CRTVU)
( which are distributed in
location or network.
Based on grid infrastructure, the platform can
help educators work in collaboration more widely by
controlling, sharing and employing resources in
virtual organization which is composed of dynamic
autonomous organizations, so they can provide
better service quality for learners and more
flexibility for the collaboration learning process
compared with the independent system serving their
own users.
Development of such a platform is motivated by
a practical problem, which is to provide learning
assessment services with collaboration for large-
scale, dynamic learners and educators group.
In the first phase, we built “Distance Learning
Assessment Grid”(LAGrid) targeting to solve the
problems as follows:
To begin with, three basic grid problems should
be resolved. The first one is its distribution. For
resources in CRTVU are distributed in the country,
LAGrid is a large-scale, distributed platform serving
the whole nation’s business process. The second
one is the dynamic and openness. CRTVU is an
organization serving learners not only from
universities but also from companies or other
organizations. The last one is the cross-organization
problem. The platform should support the cross-
organization collaborative business process. The
connected resources are from different
organizations, and the control flow and the data flow
of business process both cross organization
boundary. In a word, these three problems propose
the important requirement for building scalable grid
Secondly, there are various collaboration
patterns ranging from collaborative processes to
discrete collaborative activities. The learning
assessment task from different locations and
organizations is transparent to teachers and learners.
Users can customize a global monitoring view
anytime. Learners can access the learning resources
from various locations and organizations they need.
At the same time they can attend group learning
activities such as discussions or conferences. In
addition, managers can expand the scale of platform,
including the varieties of the business. These
applications bring higher level requirements, so we
should build a set of advanced services.
We propose the architecture to meet the
requirements of domain application as presented in
figure 1. The architecture contains five layers from
bottom to top. Firstly, at the lowest layer, the
infrastructure layer supports basic networking
environment, including computing devices, network
and networking protocols etc. Secondly, the basic
service oriented architecture layer supports the run
time environment for implementing the basic web
services related protocols such as HTTP(S)
OWL-S etc. This layer provides the elementary
connectivity, interoperation, reliability and
flexibility for the layers on top of it. Thirdly, the
grid middleware layer is the core of the architecture
where the basic grid problems such as distribution,
dynamic, openness and cross-organization are
resolved. By using SOA technology, we can design
this layer when application functionality can be
provided and consumed as sets of services. The
CSCW layer is on top of grid middleware layer to
provide collaborative work/learning services.
Lastly, the domain specific service and application
layer supports domain specific collaboration
application. New collaborative interaction services
can be published in the environment so that a
customized domain cooperation platform can be
built up.
For LAGrid, the top layer is learning assessment
service and application including learning
assessment task assignment, task submission, e-
Portfolios etc. LAGrid emphasize the grid
middleware layer and the CSCW layer. A brief
introduction of them is as follows.
3.1 Grid Middleware Layer
This layer is a crucial layer to build a grid
environment. To resolve the basic grid problems
such as distribution, dynamic, openness and cross-
organization, some relatively universal services
should be developed to provide and consume the
application functionality as sets of services. Though
the services are somewhat domain related, they are
universal to generic grid-based platform, and can be
used in other grid-based platform for e-Government,
e-Business etc. So we can categorize them as grid
middleware layer. Message Oriented Middleware
(MOM), Services Aggregator, Grid Information
Services and Reliable Bulk File Transfer Services
construct the essential components of this layer. The
main functions of them are listed below:
As a new distribution communication paradigm,
MOM, in particular its Publish/Subscribe event-
driven messaging paradigm may gradually be more
and more mature (MSMQ, 2000; Sun, 2004). MOM
supports the function of awareness events routing,
reliable transferring, and storing as the basis for
collaborative awareness in LAGrid. It has the
following functional features:
The capability of message routing based on
LAGrid’s organization ID; Support message
unicast, multicast, and broadcast; Provide the
messaging reliability mechanisms; Implement
events of publishing and subscribing functions by
Web Services technology so that the components of
MOM can be expanded flexibly.
Service Aggregator provides the function to
aggregate many services as a new service so that the
complexity of distribution, dynamic and
heterogeneous of the services can be transparent to
the users.
Grid Information Service (GIS) is a kind of
directory service (Karl, 2001). We adopt a dynamic
and scalable architecture for it. It has the following
functional features:
Describe information based on XML,
implement Web Services standards; Support
dynamic organization construction; Support the
scalability of LAGrid and the consistency in
deployment of grid runtime environment and
maintenance of grid information; Provide the
function of discovery and registration of services;
Dynamically resolve message routing address for
We develop middleware services such as Service
Management and Grid Monitor components in this
layer for reliability, stability and maintainability of
the platform. Grid Monitor is responsible for
monitoring the runtime status of grid including the
working status of grid middleware, invoking status
of Web Services instances, running status of grid
nodes etc. Service Management can support
Figure 1: LAGrid Architecture
Infrastructure Layer
Basic Service Oriented Architecture Layer
Grid Node Deployment & Management
Service Aggregator
Data Mediator
Service Composition Service Management
Replication Management
RBFT (Reliability Bulk File Transfer)
GIS (Grid Information Service)
Service Match Engine
MOM (Message Oriented Middleware)
Ontology Service
Cooperative Tools
Learning Assessment Services & Portal
Domain Specific Services
& Application Layer
Group Memory &
Collaborative Context
Task Schedule
Collaborative Awareness
management of the change of the service binding
address and the out of work of service.
Further in this layer we develop Ontology
Service, Service Match Engine and Service
Composition for the sake of new generation
knowledge-intensive cooperation environment based
on knowledge grid in the future (Cannataro, 2004).
LAGrid makes use of Ontology Service to query and
process the learning resources, services and
cooperative context. Service Match Engine matches
query conditions with services’ description based on
service ontology. Service Composition dynamically
composes services and invokes them with different
granularity. Thus the business process can be
composed automatically.
Replica management service in grid middleware
layer provides guarantee for better quality of
resource sharing, which implements functions of
transparent data transfer/copy, transparent copy
selection in grid.
Grid Node Deployment & Management is
another important component in this layer to support
the flexible scalability of grid. For example, if new
business organization joins in, a new grid node
supporting it can be created and joins in LAGrid. If
the business supported by a grid node expands, it
can be migrated to a new grid node. New grid node
integrates with original environment seamlessly in
both cases. As a result, grid nodes in LAGrid can be
expanded on users’ demand and adaptive to scale of
its business.
Now, a grid infrastructure is enacted, and it is
supported by grid middleware and is scalable.
Moreover, business service can be developed and
deployed on demand.
3.2 CSCW Layer
The cooperation work environment is the important
infrastructure built on the top of grid middleware.
In LAGrid, we mainly focus on the research of
large-scale collaboration, which includes monitoring
of cooperative process, scheduling of task in wide
area. We also consider providing team-space for
group of different sizes, where group members can
work together with groupware tools and other
business functions binding to the virtual workspace.
These groupware can support various cooperative
patterns for different cooperative tasks.
Collaborative Awareness Service provides the
facility for naturally cooperation, and it is used in
many scenarios in the cooperative environment
(Ramduny, 1998). Cooperative interactive tools also
provide communicative media in grid, and they are
categorized to common interactive tools and
domain-related cooperative tools.
Group Memory and Collaborative Context
management are used for knowledge building in
collaborative learning too (Prinz, 1993). Group
Memory is used to capture knowledge or
information within an organization and distribute it
to the workers/learners who need it (Klemke, 2000),
and context has been recognized by a wide range of
researchers as being an important concept to
consider in enhancing individual's access to
organizational information. In LAGrid, they are
used to enhance the awareness and sharing of
cooperative information. For example,
automatically deliver documents according to
context, review or reference the learning process,
assemble cooperative information about some
person, place or task etc.
Monitoring Service monitors the whole
collaborative process through a unified global view
in a real-time and on-demand manner. The
monitoring information concerns with various
statuses of users such as numbers of learners who
have or have not submitted their job, number of
teachers who have completed their tasks.
Task Scheduling Service schedules resources on
the grid nodes globally. For example, selects the
available task and distributes it to users all around
the platform. Thus the performance of the whole
grid can be improved.
3.3 Learning Assessment Services &
On one hand, Learning Assessment Layer creates
the assessment environment in grid environment to
provide the functions of creation, publishing,
processing and feedback of assessment tasks. On
the other hand, this layer makes it possible for
sharing of learning resources.
LAGrid portal is the unified entry for all grid
users. Users from different organizations who logon
it could share learning resources without knowing
where they come from.
4.1 Grid Middleware Layer
4.1.1 Message-Oriented-Middleware (MOM)
MOM consists of three main modules as presented
in Figure 2. The Core Module is composed of
message sending and receiving web services. The
Message Queue Management Module maintenances
messages in and out. The Common Service Module
is composed of message addressing service, message
routing service, security service, protocol
transforming service and application notifying
service etc. Most of the functions are implemented
as web services which can be invoked, published,
discovered and composed. As a result, the
flexibility of MOM can be enhanced by employing
such SOA-based technology in MOM design,
especially in message addressing and routing.
There are three modes of sending message which
are uni-casting, multicasting and broadcasting. The
message addressing of the three modes is through
GIS, but their mechanisms are not the same. The
message receiver for uni-casting is an organization’s
ID, but for broadcasting, it is a domain’s ID which
represents global domain or local domain. As for
multicasting, the receiver is several IDs from
different organizations or domains.
It is very important for MOM to guarantee the
reliability of message. Therefore, multi-fold
methods are used in LAGrid to support the reliability.
Multi-message-queues are employed to tackle
performance bottleneck. Error tolerance
mechanisms are considered to avoid message losing
due to application failure. Also message receipting
mechanism is used to validate message receipt.
4.1.2 Grid Information Service (GIS)
Grid Information Service of LAGrid manages four
kinds of grid entity metadata: metadata of grid
nodes; metadata of organizations, domains and
department; metadata of service interface and
metadata of services.
Since relation model better meets the demands of
the dynamic feature of grid environment (Dinda,
2000), it is used in GIS of LAGrid. We extend the
organization, domain and grid node entity based on
those defined in UDDI specification.
TModel structure is one of the most important
data structures in UDDI (UDDI, 2002). It represents
the category of entity with a unique global key
value. To improve the query efficiency of GIS,
more than ten tModels are created. These tModels
can be divided into two categories, one of which
represents the category tree of LAGrid, and the other
represents the categories of service interface. There
are three kinds of category tree in LAGrid: virtual
organization tree representing the structure of virtual
organization, organization tree representing
organizations in LAGrid and message transmitting
tree representing the logic structure of grid nodes
catering for message transmitting. When new grid
node joins in, the organization it belongs to should
be given. GIS maps the organization onto the virtual
organization tree and message transmitting tree.
Based on the mapping information, grid nodes form
different logic layered structures. The categories of
service interface include message receiving service
interface and tens of business-related service
interface. Other new service interfaces can be
further added dynamically.
Information publishing service, information
discovering service, metadata schema service and
information cache service are key services in GIS as
presented in Figure 3. GIS has close relationship
with MOM. Information changes of GIS on portal
node are synchronized to cache of every grid node.
4.1.3 Service Aggregator
This component is implemented as a compound web
service. It is designed according to a basic layered
data model that consists of three layers such as data
entity service layer, service aggregate layer and
cross-organization aggregate layer. Encapsulation
of data entity is completed in data entity service
layer; Aggregation of services from common
organization is completed in service aggregate layer;
while aggregation of services from different
organizations is completed in cross-organization
aggregate layer.
When an aggregating request comes in, services
are discovered and selected at run time invoking grid
Message Locating
User Define
Message Routing
Protocol Translating
Message Seuring
User Define
Public Services
User Define
Web Service Interface
Message Oriented Middleware
Figure 2: Architecture of MOM
information services. Then they are filtered based
on aggregating policy. Before invoking it, service
consumer negotiates security policy with the service
providers using LAGrid security component. Then
the request is distributed to service providers on
different grid nodes. The instances of services are
generated dynamically and are invoked in parallel.
At the end of aggregating process, results returned
from service providers would be transformed and
In LAGrid, data sources such as learners’ score,
users’ attachment in e-portfolio from different
organizations need to collected real time. This
brings more complexities in aggregating them.
Figure 3 shows how GIS and Service Aggregator are
used in aggregating such business data distributed in
wide-area network environment.
4.2 CSCW Layer
CSCW layer of LAGrid provides collaborative
awareness service, collaborative context service,
monitor service, task scheduling service and so on,
based on the grid middleware layer.
Collaborative awareness service consists of three
parts, which are event generation component, event
processing component and awareness mediation
Event generation component is to capture and
publish related events. Events in LAGrid can be
categorized into grid metadata synchronization
events, grid maintenance events and business events.
Upon update of grid nodes or services,
synchronization events are generated. Grid
maintenance events are caused by grid exception.
Business events are triggered by time, users or some
specific activity.
Event routing, transferring, and storing are
implemented based on MOM. Event processing
component on every grid node is an independent
windows process. Those validated events can be
presented at online message notification interface, or
sent to email server or processed automatically by
specific agent.
Awareness mediation component functions as an
enhanced module to support personalized event
subscription and redirection. Users of LAGrid can
subscribe interested events and redirect the events to
suitable event processing component. For example,
teachers may subscribe exercise finishing event of
specific student and get the notification by email.
Collaborative context service consists of two
parts which are group memory service and context
management service. Group memory service acts as
a store of collaborative information relevant to a
user performing a task in LAGrid. Context
management service models the collaborative
context from various dimensions and provides
means of query/reasoning of collaborative context.
Collaborative context service is implemented
based on Ontology Service, Service Match Engine
of grid middleware. Metadata of collaborative tools
such as server’s port, task type supported, group size
permitted etc. is encapsulated as collaborative
service. Collaborative service publishing and
discovering are implemented based on top-level
collaborative ontology and collaborative service
ontology model. Details will be discussed in the
other papers.
Monitor Service is implemented based on
Service Aggregator. Task status on every grid nodes
is encapsulated as web service, and a unified
monitoring view is generated based on Service
Aggregator. Task Scheduling Service implements a
global scheduling algorithm.
Email, online message tool are integrated in this
layer and new interaction tools such as threaded
Figure 3: The Mechanism Scheme of Data Aggregating
Information Service
wsdl wsdl
- Data Input & Output & format
- Methods to get data
- Security Requirements
- message exchange protocol
- node
- service interface description
- service location
- topology
- node (address, localScope, etc.)
- service interface description
- service location
- VO
- Organization Entity
Service Aggregator
Metadata Schema
InfoService Cache
discussion tools and whiteboard will be integrated in
plug-and-play manner.
4.3 Learning Assessment Services &
Learning Assessment Layer provides almost all
functions in assessment task process on top of other
layers. They are task creation, publishing,
delivering, retrieving, submitting, and statistics etc.
Figure 4 shows the main assessment actions LAGrid
Sharing of assessment results and learning
resources are implemented. The sharing process can
be divided into three phases. In the phase of
resources publishing, when resource providers
publish resources, the metadata of resources are
generated and encapsulated in a message. Then it is
broadcasted on the whole system. In the phase of
resources delivering, metadata of resources be
subscribed are delivered through MOM. Service
Aggregator of grid middleware is used in the phase
of resources query.
In addition, various business messages
notification and feedback are implemented based on
Publish/Subscribe mechanisms.
Collaborative learning is also supported through
team-space where various collaborative tools based
on services provided by CSCW layer.
The Single Sign On (SSO) component
implements LAGrid portal by providing grid user
security, grid node security and web services
security in an unified manner.
Large-scale cross-organization e-Learning
environment is also investigated by some national
projects. The National Grid for Learning (NGfL, is a Government initiative
to help learners and educators in the UK benefit
from information and communications technology
(ICT). NGfL collects many website links into a
single portal by hyperlink. It is based on traditional
web technologies rather than service-oriented
architecture. NGfL integrates website links rather
than data or function. Although learners and
educators can access resources on different websites,
data and processes of different organizations remain
independent and can’t be orchestrated efficiently.
Some approaches such as Access Grid
(, which aims to support
wide area, real-time, and computer-mediated
communication in large-scale distributed conference
and training, and ELENA (http://www.elena-, which focuses on providing
personalization, openness, and interoperability in the
context of smart spaces for learning by using
semantic web technology and adaptive learning
management technology have different focus with
LAGrid. GECSR(Grid- Enabled Collaboratory for
Scientific Research) (GECSR, 2004), which aims at
building new generation of global-scale
collaboration infrastructure and focuses on
integration of distributed collaborative tools, agent
based monitoring and decision support system
according to a proposal submitted to NSF, is
somewhat similar to our work. However, further
development progress and more detailed
technologies have not been reported yet.
Task Checking
Task Dispatching
Global Monitoring
Resource Sharing
Figure 4: Learning Assessment Actions Supported by LAGrid
Grid-based collaborative environment architecture
of domain oriented application is proposed based on
research on combination of grid computing and
CSCW. Grid middleware and CSCW services are
developed which are MOM, GIS, Service
Aggregator, Collaborative Awareness Service,
Collaborative Context Service, Monitor Service and
Task Scheduling Service etc. Collaborative process
in distributed, dynamic, open, cross-organization
environment is supported by those services.
Furthermore advanced collaborative learning service
can be developed for the learning assessment task in
e-Learning. Therefore, our work fills the existing
gap between current cooperative platform and
complex, cross-organization infrastructure.
Services of LAGrid have been deployed in five
cities ( supporting assessment
task of three courses for CRTVU. The next phase of
the projectGrid-based Collaborative Platform for
e-Learning will aim to promote perfect
development of grid middleware and CSCW
services. More flexible CSCW services will be
delivered together with the more robust and flexible
grid middleware components. Various collaborative
resources can be reused and composed according to
context. Various cooperation-intensive groups in
different sizes may coexist with each other orderly
and may be evolved in a natural way.
Our project is supported by the National Natural
Science Foundation of China under grant No.
90412009. The authors would like to acknowledge
the other members of the LAGrid team for their
GECSR, 2004, ‘ITR: A Global Grid-Enabled
Collaboratory for Scientific Research’,
M. Atkinson, D. DeRoure, A. Dunlop, 2004, ‘Web Service
Grid: An Evolutionary Approach’. UKeS-2004-05,
TAG revised version, <>
H. Cowie and J. Ruddick, 1988. Cooperative Group Work,
an overview, BP Educational Service, Sheffield
Lopez, P. G. & Molla, R. R., 2001. ANTS : a new
Collaborative Learning Framework. In
EuroCSCL’2001, European Conference on Computer
Supported Collaborative Learning.
Foster, I., C. Kesselman, S. Tuecke., 2001. The Anatomy
of the Grid : Enabling Scalable Virtual Organizations,
In t. J. of Supercomputer Applications and High
Performance Computing ,15(3) 2001,pp.200-222
Foster, I., C. Kesselman, J. Nick, S. Tuecke., 2002. The
Physiology of the Grid: Open Grid Services
Architecture for Distributed Systems Integration, In
Proc. 4th Global Grid Forum (GGF) Workshop 2002,
Microsoft Message Queue (MSMQ) Center, 2000.
Sun, Java Message Service Specification, 2004.
Karl Czajkowski, Steven Fitzgerald,Ian Foster,Carl
Kesselman, 2001. Grid Information Services for
Distributed Resource Sharing, Proc. 10th IEEE
International Symposium on High-Performance
Distributed Computing (HPDC-10), IEEE Press.
M. Cannataro and D. Talia, 2004. "Semantics and
Knowledge Grids: Building the Next-Generation
Grid," IEEE Intelligent Systems, vol. 19, pp. 56-63.
Ramduny, D., A. Dix and T. Rodden, 1998. Exploring the
Design Space for Notification Servers”, in Proc. of
CSCW’98 : ACM Conference on Computer Supported
Cooperative Work, Seatle, ACM Press, pp. 227-236.
Klemke, R., 2000. Context Framework - An Open
Approach to Enhance Organisational Memory
Systems with Context Modeling Techniques. In
Proceedings of the International Conference on
Practical Aspects of Knowledge Management - PAKM
2000 (Oct. 30-31, Basel, Switzerland).
W. Prinz., 1993. TOSCA – Providing organizational
information to CSCW Applications, In Proceedings of
Int’l Conference on Computer supported Cooperative
Work (ECSCW’93) Kluwer Press.
P. Dinda and B. Plale, 2000. A Unified Relational
Approach to Grid Information Services, Grid Forum
Informational Draft GWD-GIS-012-1.
UDDI Version 2.03 Data Structure Reference UDDI
Committee Specification, 19 July 2002,