--------------------------------------------------------------------------
           A Report on Professor Sushil K. Prasad's Visit
              The University of Melbourne, Australia
                        July-Aug 2006 
--------------------------------------------------------------------------

Professor Sushil K. Prasad of Georgia State University, Atlanta, USA, was
funded for a two-month research visit to the University of Melbourne,
Australia, during summer 2006 to collaborate with Prof. Buyya on Grid and
Mobile Computing research, with Prof. Palaniswami on Sensor Networks, and
with NICTA.  The outcomes of the visit are as follows:

1. Prof. Prasad gave a series of talks to the Grid Computing group at Univ.
of Melbourne on Parallel Algorithms and Data Structures, and on Middlewares.  
The final talk was on an open NICTA/UMLB colloquium entitled "SyD: A
Middleware for Collaborative Applications over Small Heterogeneous Devices
and for Distributed Biological Workflows over Web Services."

2. A joint proposal to National Science foundation was submitted entitled
"PIRE: US - Australia - Canada Partnership in Developing Middleware
Technologies for Enabling Mobile Grid" with Prof. Buyya and Prof. Palani.  
This also resulted in international collaborations with sensornet group at
Oak Ridge National Lab and Acenet group at University of New Brunswick,
Canada.

There are good reasons for integrating the handheld devices, such as cell
phones and personal digital assistants (PDAs), and the myriad of sensor
networks into the current computational and data grid infrastructure: (i)
these devices are increasingly getting more capable in processing power,
memory, and battery life, (ii) are getting better connected, and (iii) are
proliferating.  Harnessing their capabilities holds the key to a range of
applications from homeland security to e-science.  Prof. Prasad has designed
and implemented a middleware for handheld and other devices to enable
seamless embedded software development and their deployment on a
heterogeneous set of devices across multiple networking protocols and data
formats.  They have also developed and prototyped a preliminary system which
allows users to develop and execute simple workflows over actual web services
deployed over the Internet, while the workflow is launched and monitored
through a mobile PDA.

Prof. Buyya and Palani have developed robust middlewares for distributed
computation and workflows over the grid infrastructure, and are developing a
middleware for interfacing with sensor networks.
  This PIRE project proposed to leverage of these complementary works, and
research into developing a mobile grid infrastructure which seamlessly allows
(i) employing mobile devices as a window to the global grid, and (ii)
extending the global grid to the mobile devices, wherein these devices act as
servers for data (stored and sensed) and synergistically collaborate among
themselves and with core grid for greatly enhanced utility.

3.  Prof. Prasad co-advised a Ph.D. student of Prof. Buyya's lab, Anthony
Solisto, on advance grid scheduling techniques, resulting in a joint paper
entitled "GarQ: An Efficient Data Structure for Advance Reservations in Grid
Computing" submitted to Intl.Conf. on Parallel Processing (ICPP).  A follow
up journal submission is planned.

In Grid systems, users may require assurance for completing their jobs on
shared resources. Such guarantees can only be provided by reserving resources
in advance. However, if many reservation requests arrive at a resource
simultaneously, the overhead of providing such service will be significant.
An efficient data structure for managing these reservations plays an
important role in order to minimize the time complexity for searching
available resources, adding new requests, and deleting existing reservations.
Several data structures have already been proposed for related scheduling
problems, e.g. for admission control in network bandwidth reservation,
including Linked Lists and Segment Tree.

In this paper, they first describe suitably modified versions of the Linked
List and Segment Tree data structures in capable of dealing with advance
reservations in computational Grids. For Segment Tree, this entailed
developing a new algorithm for finding a free interval closest to the
requested reservation. Next, it describes how these operations can be
performed on Calendar Queue, a data structure commonly employed for discrete
event simulations. It then proposes a Grid advanced reservation Queue (GarQ),
which is a new data structure based on Calendar Queue and Segment Tree, that
improves some weaknesses of the aforementioned data structures. We
demonstrate the superiority of the proposed structure by conducting a
detailed performance evaluation on real workload traces.

4.  A project has been started to create a light-weight implementation of
gridbus broker to enable it to interface with web services.  The students
involved are Chad Christopher and Yan Chen from Prasad Lab and Krishna
Nandiminti and Srikumar from Prof.Buyya's lab. The main idea was to create a
stack of lightweight componentswhere the scheduler was to be replaced by a
'matcher' and calls to the execution framework removed.  The entire
architecture has been laid out.

5. Some joint initial exploration has been carried out in utilizing SVM
techniques developed by Prof. Palani's lab into formulating distributed
algorithms for extending lifetime of sensor networks. Also, fleet application
of SyD middleware is planned to be leveraged for Prof. Palani's mobile
inventory and logistics framework.

----------------------------------------------------------------------------