Environment-aware Mobile Programs 

          Anurag Acharya, M. Ranganathan, Shamik Sharma, Joel Saltz
		      Department of Computer Science
     	        University of Maryland, College Park 20742

Mobile programs are expected to operate in an environment where the
availability of resources changes dynamically and is not under the
control of the user. For example, network characteristics change based
on traffic; server availability changes based on usage and
administrative decisions. To be effective and efficient in such an
environment, mobile programs need to be able to monitor the level and
quality of resources available and to adapt to significant changes.

As a part of the "Green" project (on environment-aware programs), we
are seeking answers to the following four questions.

1. How rapidly does the environment change and how stable the changes
   are?  Also, by how much does the environment have to change for it
   to significantly impact the operation of a mobile program?

2. What is the cost of monitoring the environment? 

3. What is the benefit of adapting to environment changes?

4. How should mobile programs be structured to be able to adapt
   to changes in their environment?

We are addressing these questions in an empirical manner - by building
and studying environment-aware mobile programs. We have developed an
infrastructure for developing adaptive mobile programs and for
distributed resource monitoring. We are currently in the process of
implementing and studying a set of adaptive mobile programs.

Based on early experience with a network-latency-sensitive application
and a study of end-to-end network latency, we have developed three
working hypotheses.  First, a resource-aware placement of components
of a distributed program can provide significant performance gains
over a resource-oblivious placement. Second, effective mobility
decisions can be based on coarse-grained monitoring. Finally, a simple
and cheap distributed resource monitoring scheme can provide
sufficient information for effective mobility decisions.  We found
that latency-sensitive programs that run for at most an hour or so
gain little from continuous monitoring and that most of the gains for
these programs can be achieved by on-demand determination of resource
levels. On-demand monitoring would also be suitable for resources that
are more expensive to monitor, for example, network bandwidth.

We are taking two directions in our current research. First, we are
working on developing cheap-to-compute measures for other resources,
in particular, network bandwidth and server availability.  Second, we
are developing several other mobile programs.  Our target applications
include (1) resource-aware queries that span multiple medical
databases (this effort is in collaboration with researchers at the
Johns Hopkins Medical School), (2) resource-aware pre-fetching for web
clients and (3) custom combination of periodically generated
large-volume datasets (such as weather information).