WorkshopIndex
Add your thoughts about topics for the workshop here:- I'd like to talk about how to design hugely large systems using genetic algorithms, genetic programming, or something like that. rpg
- Is the (first) era of Software Engineering over? rpg
I would like to talk about what role Simplicity@ (of software) plays in ULS or its parts. PeterSommerlad
- It would be nice to discuss the antinomy, if any, of modularitystructure, and evolutionemergence (in a biological sense). EricTanter
- An autopoietic system is one that produces its own boundaries as part of its process. What does it mean for a software? what are its boundaries? EricTanter
- Can domain-specific modeling languages (!DSMLs) help in the integration of ULS systems? What advances in the tools are necessary to support integration at the DSML level? Krishnakumar
- Will a model (usually an XML file conformant to a standard schema, or a proprietary format) in the current technologies retain the same characteristics (single file loaded into an IDE) in the context of ULS systems? How can models be effectively shared among developers? Krishnakumar
- Are existing modeling approaches even appropriate for ULS's? Leaving aside the model representation issues inherent in Krishnakumar's question, are there changes or additions to standard modeling representations that must be created for ULS's? How do we scale not just the artifacts but the processes as well? KyleBrown
- Is it possible to create mega software product line architectures by combining the software product line architectures of individual participants (federations) in the context of ULS systems? Will !DSMLs help with this? What will it take to attempt such integration? Krishnakumar
- I'd like to talk about what makes software robust as opposed to brittle, and how we might make the interfaces between modulescomponentsservices be richer and more forgiving. RonGoldman
- I'd like us to consider when we can ignore efficiency and the consequences of free cpu cycles & free programmer time. RonGoldman
- In a ULS system, FailureIsConstant. It's not a matter of what to do when the system fails, but a matter of "hundreds of components are failing all the time" how can we design and implement software to live well with that? Which research needs to be done to support that? Do we need new Patterns? Services? Tools? Languages? Middleware? FabioKon.
- We'd like to talk about the design of decentralized component deployment and recovery algorithms and adaptive middleware platforms to support high-resilience ULS system development and deployment against a wide spectrum of disturbances, including failures of system parts, as well as physical and cyber attacks. YuanXue
- Considering the number of autonomous systems and how human are part of the ULS, we should probably look into Ontology to facilitate gathering, assimilating, and reasoning information from multiple systems/domains consistently. NanborWang
- Are ULS's BornOrMade? In other words, can a large-scale system grow into a ULS or must a ULS be designed from the ground up as such? KyleBrown
- Alongside with FailureIsConstant, we also have ConfigurationChangesAreConstant - as at any given time there will be components who has just been updated. The NASA paradigm of TestWhatYouFlyAndFlyWhatYouTest seems to mean either keeping your system frozen until you have enough resources for a very extensive regression test or looking the other way (see an example here http:/catless.ncl.ac.ukRisks/24.33.html#subj1.1) How do we cope with this rate of change? What do we need in order to meet system-wide goals such as availability and reliability?AmirRaveh
- ULS pose huge challenges on software understanding, in terms of several kinds of knowledge, which is crucial for effective software reuse. My biggest motivation related with UnderstandingULS, are the challenges ULS imply in terms of techniques to help on understand them better, with particular interest in terms of design knowledge. AdemarAguiar
- We seem to have run into new and unexpected problems with each order of magnitude increase in program/system size. How can we test at a reduced scale in a convincing manner that will let us build up to the full scale and encounter the unexpected problems before we deploy at full scale. JohnDougan
- Further to the testing question, what problems are suitable for using as model problems for research and testing? What will be the Drosophilia of Ultra Large Scale systems? JohnDougan
- It's not known to many people, but there has been a bunch of work done on mechanism design, particularly market mechanisms, as applied to computation. The people in question call their approach AgoricComputation and have done a fair amount of work on software to support this approach. What can we steal from their work? JohnDougan
- James P. Hogan's novel The Two Faces of Tomorrow [(full text here) http:/www.webscription.net10.1125Baen0671878484/0671878484.htm] provides a dramatized look at issues involved in ULS systems. Summary: The planetary computation network has started to exhibit unexpected and dangerous behavior. Dr. Ray Dyer's research team has a possible fix but will the cure be worse than the disease? They need to test the new system, but how will they do it? JohnDougan