Alva L. Couch
>
Research
( couch@cs.tufts.edu )
Quick links:
Contact
Teaching
Publications
Research
Software
Personal
Arts
Archives
This page does not describe my prior publications or software
distributions; just ongoing work prior to any publication. For details
on publications or software, use these links:
Research
There are currently two inter-related themes in my research:
network and system operations and management, and geo-informatics.
The former studies how to effectively manage and support
high-reliability systems and services, while the latter applies system
management principles to big data science. I will describe the latter
before the former, from new work to old.
Data management
The field of data management concerns how scientific and technical
data is managed, stored, protected, and retrieved. My work in this
field has centered upon the problems that arise in geo-informatics in general,
and the problems of hydrologic information systems in particular.
Problems upon which I have worked include:
- Defining a useful and acceptable data lifecycle, from creation, through use, to deprecation and eventual removal.
- Tradeoffs in choosing a storage medium for big data.
- Tradeoffs between local and cloud-based data storage.
- New data discovery mechanisms that enable data discovery by scientists outside the scientific expertise of the original data gatherers.
- Use of metadata ontologies to enable data discovery.
- Use of telescoping search interfaces to avoid information overload during the discovery process.
At present, I am the proposed interim director for a new Water Data
Center to operate under the auspices of the Consortium of Universities
for the Advancement of Hydrologic Science, Inc (CUAHSI). In this
role, I will study the problems and processes of data curation for a
large federated data delivery system, and will move parts of that
system into the Microsoft Azure Cloud.
Network and System Operations and Management
I study enabling technologies for system administration and autonomic computing:
- System administration refers to the task of building and maintaining networks of computers to achieve some predetermined business or research objectives.
- Autonomic computing refers to the practice of replacing human system administrators with programs or agents that maintain network function automatically.
Thus system administration and autonomic computing differ mainly in whether computers or human beings are managing the network.
I have studied many aspects of system administration, including:
- Cooperative management: tools and technologies that allow system administrators to cooperate and collaborage in maintaining systems (e.g., SLINK (1996)).
- Troubleshooting: determining why a computer network is not functioning properly and taking corrective actions.
- Cost modeling: predicting the total cost of management by modeling the cost of management and contingencies (e.g., network outages).
- Dependency analysis: determining how interactions between components in a network influence the function of a network, and predicting the impact of changes to the network.
- Configuration management: creating and updating the definition of what a network should do, which is often called its "configuration."
- Service management: insuring that services required for a functional network are present and functional., e.g., address management, web service, electronic mail.
- Performance analysis: determining and correcting delays in providing network services.
My key contributions to the field of system administration include:
- Tools that enable community-based system administration (1996).
- Analysis of the relationship between troubleshooting methods and cost of ownership (2000,2005).
- Simple, useful tools for dependency analysis of complex systems (2001).
- An innovative approach to depicting network function via sound (2001).
- A theoretical model of configuration management (2003,2006).
- A definitive discussion of the configuration management problem (2006).
- Definition and analysis of statistical notions of system dependency (2008).
Recently, I have shifted my attention toward technologies that enable collaboration between human and non-human (autonomic) management agents. There are two prevalent approaches to autonomic computing:
- Closed-loop modeling: controlling a system based upon a complete model of its reactions to changes in its environment.
- Immunological approaches: management programs that detect problems and take action only when problems arise. The management agent "immunizes" the system against observed problems.
Together with a small number of others, I am studying the immunological approach. This approach has several aspects that do not arise in the control-centric version of autonomics:
- Convergent operators: computer programs that - when repeatedly applied - maintain some desirable state of operation for a network.
- Closures: closed and predictable components of an otherwise open and unpredictable network.
- Promise theory: a variant of logic that models voluntary cooperation between autonomous agents.
- Emergent behavior: a result of a management process that is only indirectly related to the details of the process.
My contributions to autonomic computing include:
- Definition of the concept of a closure, with theoretical implications and proof-of-concept prototypes (2003, 2005, 2006).
- Algebraic properties of convergent operators (2003).
- Analysis of relationships between control-centric autonomics and convergent operator theory (2006).
- Definition of the concept of an exterior type for a promise body, as an extension of Burgess' promise theory (2007).
- Analysis of emergent behavior of systems of convergent operators (2008).
- Convergent operators for resource management (2009).
Alva L. Couch
>
Research
( couch@cs.tufts.edu )
Quick links:
Contact
Teaching
Publications
Research
Software
Personal
Arts
Archives