Summary
Quality-of-service (QoS) enabled publish/subscribe (pub/sub) distributed,
real-time, and
embedded (DRE) systems often need to satisfy tradeoffs between
multiple (often contentious) QoS demands e.g., security, timeliness,
and reliability. These systems also need to regulate and adapt to
(dis)continuous changes in runtime environments to support features
such as online prognostics, dependable online upgrades, availability
of mission critical tasks, and dynamic resource management.
In particular, as an operating environment changes, QoS mechanisms that
were once adequate might no longer be effective thus failing to support
specified QoS. Building on the work with
FLEXMAT
we are focusing on the particular
QoS mechanisms of transport protocols. We are conducting research so
that pub/sub middleware can autonomically adapt transport protocols
to support QoS in dynamic environments. Our solution approach called
ADAptive Middleware And Network Transports (ADAMANT) integrates and
enhances QoS-enabled pub/sub middleware, a flexible
network transport protocol framework, monitoring infrastructure,
machine learning techniques, and a controller to support autonomic
adaptation of QoS mechanisms for pub/sub DRE system in dynamic
environments.
We use the
Data Distribution Service
(DDS) as our initial pub/sub middleware since it provides
fine-grained QoS policy control. DDS is a robust pub/sub API
standardized by
the Object Management Group (OMG).
We have integrated and enhanced the
OpenDDS and
OpenSplice open source
DDS implementations with an Adaptive Network Transports (ANT) framework
and are currently focusing on the
OpenSplice implementation due to its extensive
commercial grade support of the DDS specification.
We are using the ANT framework to provide
adaptability of transport protocols and protocol parameters. ANT is a
multicast-based transport protocol framework that supports
composable modules with particular properties (e.g., acknowledgment-based
reliability, negative acknowledgment based reliability, lateral
error correction, XOR data encoding). The modules in ANT can be
reconfigured at run-time to provide transport protocol adaptability.
Status and Current Work
We are now integrating the various pieces of ADAMANT (as shown below) to
implement an
ADAMANT prototype for dynamic environments. We will then conduct
experiments and collect performance data (e.g., measure the time taken
from a change in the environment to a change in protocols).
|
The subversion URL for ADAMANT is located at: svn://svn.dre.vanderbilt.edu/DOC/ADAMANT.
ADAMANT Publications, Presentations, and Posters
-
Joe Hoffert, Aniruddha Gokhale, and Douglas Schmidt,
"Timely Autonomic Adaptation of Publish/Subscribe Middleware in Dynamic Environments",
International Journal of Adaptive, Resilient and Autonomic Systems;
Vol 2, Issue 4 (2011), pp. 1-24.
-
Joe Hoffert, Douglas Schmidt, and Aniruddha Gokhale,
"Evaluating Timeliness and Accuracy Trade-offs of Supervised Machine Learning for Adapting Enterprise DRE Systems in Dynamic Environments",
International Journal of Computational Intelligence Systems, Volume 4,
Issue 5 (2011), pp. 806-816.
-
Joe Hoffert, Daniel Mack, and Douglas Schmidt,
"Integrating Machine Learning Techniques
to Adapt Protocols for QoS-enabled Distributed Real-time and Embedded
Publish/Subscribe Middleware",
International Journal of Network Protocols and
Algorithms (NPA): Special Issue on Data Dissemination for Large-scale
Complex Critical Infrastructures,
Vol 2, No 3 (2010)
-
Joe Hoffert, Douglas Schmidt, and Aniruddha Gokhale,
"Adapting Distributed Real-time
and Embedded Publish/Subscribe Middleware for Cloud-Computing
Environments",
Proceedings of the ACM/IFIP/USENIX 11th International Middleware Conference
(Middleware 2010),
Bangalore, India, November 2010
-
Joe Hoffert and Douglas Schmidt,
"Evaluating Supervised Machine
Learning for Adapting Enterprise DRE Systems",
Proceedings of the 2010 International Symposium on
Intelligence Information Processing and Trusted Computing (IPTC 2010),
Huanggang, China, October 2010
-
Joe Hoffert, Douglas Schmidt, and Annirudha Gokhale,
"Adapting and Evaluating Distributed
Real-time and Embedded Systems in Dynamic Environments",
The 1st International Workshop on
Data Dissemination for Large scale Complex Critical Infrastructures
(DD4LCCI 2010),
Valencia, Spain, April 2010
-
Joseph W. Hoffert, Daniel Mack, and Douglas Schmidt,
"Using Machine Learning to Maintain Pub/Sub
System QoS in Dynamic Environments",
8th Workshop on Adaptive and Reflective Middleware (ARM) 2009,
Urbana Champaign, IL, December 2009
-
Joseph W. Hoffert and Douglas Schmidt,
"Maintaining QoS for
Publish/Subscribe Middleware in Dynamic Environments",
Fast abstract for the 3rd ACM International Conference on Distributed
Event-Based Systems (DEBS '09), Nashville, TN, USA, July 2009
ADAMANT Doctoral Consortia
-
Joe Hoffert,
"Evaluating and Adapting
QoS for Distributed Real-time & Embedded Systems in Dynamic Environments",
EuroSys 2010 Conference, doctoral symposium poster sesssion, Paris, France,
April 2010
-
Joseph W. Hoffert (Adviser: Douglas Schmidt),
"Maintaining Publish/Subscribe
Middleware QoS in Dynamic Environments",
Doctoral and PhD Workshop for the 3rd ACM International Conference on
Distributed Event-Based Systems (DEBS '09), Nashville, TN, USA, July 2009
-
Joseph W. Hoffert (Adviser: Douglas Schmidt),
"Maintaining QoS for
Publish/Subscribe Middleware via Autonomic Adaptation",
Doctoral Consortium for the 6th International Conference on Autonomic
Computing and Communications, Barcelona, Spain, June 2009