Complex software systems today can be characterized by distribution,
heterogeneity, and changing resource requirements and capacities.
These attributes make static systems unsuitable for a wide range of tasks
that need high performance, or are long-lived.
In order to achieve high performance in such environments for more than a
short period of time, systems need to dynamically adapt to changing
resource capacities and application requirements.
We are designing and building Active Harmony, a software
architecture that supports distributed execution of computational objects
in such environments through the following innovations:
- dynamic execution environment: Dynamic adaptation to network and
resource capacities, both when computational objects are created, and when
application requirements or resource capacities change.
Active Harmony will attempt to maximize data affinity and load balancing
through intelligent resource allocation and object migration.
- automatic application adaptation: A framework that
permits runtime adaptation of algorithms, data distribution, and load
balancing.
Active Harmony will export a detailed metric interface to
applications, allowing them to access processor, network, and operating
system parameters.
Applications export tuning options to the system, which can then automatically
optimize resource allocation.
Measurement and tuning can therefore become first class objects in the
programming model.
Programmers can write applications that include ways to adapt computation to
observed performance and changing conditions.
- shared-data interfaces: Active Harmony will support shared-memory
semantics among computational objects regardless of location, allowing both
peer-to-peer and client-server computations to exploit the simplified
programming model and fine-grained sharing permitted by traditional
shared-memory environments.
Innovations include support for heterogeneity of both data and
program code, a multi-level security scheme that adapts data and code
interfaces to the degree of trust between computational objects, and
support for the dynamic execution environment.
The unique aspect of the Active Harmony work is the emphasis on adapting to
heterogeneous and changing environments.
Other researchers have studied some of the constituent issues that we plan
to address.
Our emphasis is on the inter-relationships between objects in the
system.
The primary result of this research will be an infrastructure and a set of
algorithms that permit global resource optimization under changing
conditions.
|
|
|
2009 Papers-
A Scalable Autotuning Framework for Compiler Optimization
Ananta Tiwari, Chun Chen, Jacqueline Chame, Mary Hall, Jeffrey K. Hollingsworth
IPDPS 2009, Rome, May 2009.
(Abstract,  )
-
Tuning Parallel Applications in Parallel
Ananta Tiwari, Vahid Tabatabaee, Jeffrey K. Hollingsworth
Parallel Computing, (to appear) 2009.
(Abstract)
2008 Papers-
PERI Auto-Tuning
David Bailey, Jacqueline Chame, Chun Chen, Jack Dongarra, Mary Hall, Jeffrey K. Hollingsworth, Paul Hovland, Shirely Moore, Keith Seymour, Jaewook Shin, Ananta Tiwari, Sam Williams, Haihang You
Journal of Physics: Conference Series 125 (2008), Nov. 2008.
(Abstract, )
2007 Papers-
Automatic Software Interference Detection in Parallel Applications
Vahid Tabatabaee, Jeffrey K. Hollingsworth
SC'07, Reno, NV, Nov. 2007.
(Abstract, )
2006 Papers-
A Case Study Using Automatic Performance Tuning for Large-Scale Scientific Programs
I-Hsin Chung, Jeffrey K. Hollingsworth
International Symposium on High Performance Distributed Computing (HPDC), Paris, June 2006.
(Abstract, )
2005 Papers-
Parallel Parameter Tuning for Applications with Performance Variability
Vahid Tabatabaee, Ananta Tiwari, Jeffrey K. Hollingsworth
SC'05, Seattle WA, November 2005.
(Abstract, )
2004 Papers-
Using Information from Prior Runs to Improve Automated Tuning Systems
I-Hsin Chung, Jeffrey K. Hollingsworth
Proceedings of SuperComputing, November 2004.
(Abstract, )
-
Automated Cluster-Based Web Service Performance Tuning
I-Hsin Chung, Jeffrey K. Hollingsworth
Proceedings of IEEE Conference on High Performance Distributed Computing (HPDC), June 2004.
(Abstract, )
-
Towards Automatic Performance Tuning
I-Hsin Chung
Ph.D. Dissertation, University of Maryland, December 2004.
(Abstract, )
2003 Papers-
Runtime Selection Among Different API Implementations
I-Hsin Chung, Jeffrey K. Hollingsworth
Parallel Processing Letters Vol. 13, No. 2, June 2003.
(Abstract)
2002 Papers-
Active Harmony: Towards Automated Performance Tuning
Cristian Tapus, I-Hsin Chung, Jeffrey K. Hollingsworth
Proceedings of SuperComputing, November 2002.
(Abstract, )
2001 Papers-
Exploiting Idle Cycles In Networks Of Workstations
Kyung Dong Ryu
Ph.D. Disseration, University of Maryland, August 2001.
(Abstract, )
1999 Papers-
Exposing Application Alternatives
Peter Keleher, Jeffrey K. Hollingsworth, Dejan Perkovic
International Conference on Distributed Systems (ICDCS), June 1999.
(Abstract, )
-
Mechanisms and Policies for Supporting Fine-Grained Cycle Stealing
Kyung Dong Ryu, Jeffrey K. Hollingsworth, Peter Keleher
Internation Conference on Supercomputing (ICS), June 1999.
(Abstract, )
-
Prediction and Adaptation in Active Harmony
Jeffrey K. Hollingsworth, Peter Keleher
Cluster Computing, 2 (1999), 1999.
(Abstract)
1998 Papers-
Prediction and Adaptation in Active Harmony
Jeffrey K. Hollingsworth, Peter Keleher
IEEE International Symposium on High Performance Distributed Computing (HPDC), July 1998.
(Abstract, )
|
|
