HR: 09:15h
AN: H51J-06 INVITED [Abstracts]
TI: Autonomic Computing Paradigm For Large Scale Scientific And Engineering Applications
AU: * Hariri, S
EM: hariri@ece.arizona.edu
AF: The University of Arizona, Electrical and Computer Engineering Department,
1230 E. Speedway Blvd, Tucson, AZ 85721
United States
AU: Yang, J
EM: jm_yang@ece.arizona.edu
AF: The University of Arizona, Electrical and Computer Engineering Department,
1230 E. Speedway Blvd, Tucson, AZ 85721
United States
AU: Zhang, Y
EM: zhang@ece.arizona.edu
AF: The University of Arizona, Electrical and Computer Engineering Department,
1230 E. Speedway Blvd, Tucson, AZ 85721
United States
AB:
Large-scale distributed scientific applications are highly adaptive and heterogeneous in terms of their computational
requirements. The computational complexity associated with each computational region or domain varies continuously and
dramatically both in space and time throughout the whole life cycle of the application execution. Furthermore, the underlying
distributed computing environment is similarly complex and dynamic in the availabilities and capacities of the computing
resources. These challenges combined together make the current paradigms, which are based on passive components and static
compositions, ineffectual. Autonomic Computing paradigm is an approach that efficiently addresses the complexity and dynamism
of large scale scientific and engineering applications and realizes the self-management of these applications.
In this presentation, we present an Autonomic Runtime Manager (ARM) that supports the development of autonomic applications.
The ARM includes two modules: online monitoring and analysis module and autonomic planning and scheduling module. The ARM
behaves as a closed-loop control system that dynamically controls and manages the execution of the applications at runtime.
It regularly senses the state changes of both the applications and the underlying computing resources. It then uses these
runtime information and prior knowledge about the application behavior and its physics to identify the appropriate solution
methods as well as the required computing and storage resources. Consequently this approach enables us to develop autonomic
applications, which are capable of self-management and self-optimization. We have developed and implemented the autonomic
computing paradigms for several large scale applications such as wild fire simulations, simulations of flow through variably
saturated geologic formations, and life sciences.
The distributed wildfire simulation models the wildfire spread behavior by considering such factors as fuel characteristics
and atmosphere conditions. It is a highly adaptive application since the large difference in computational complexities
between "burning" and "unburned" regions leads to substantial load imbalance at runtime. The VSAFT2D (Variable Saturated
Aquifer Flow and Transport) application is a 2D finite element application designed for the simulation of water flow and
chemical transport through variably saturated porous media. The life science application involves the analysis of excessively
large microarray experiments for gene expressions. This application is an I/O intensive application which iteratively
estimates the parameters of a linear model for gene expressions by randomly sampling data from a huge data set.
DE: 0599 General or miscellaneous
SC: Hydrology [H]
MN: Fall Meeting 2005