 |
CBET
Discussion - part 4
Should there be a CBET community aimed at CI? If so, how should it be formed and organized to be a player in major CI initiatives, eg Teragrid?
- If there is a CBET community, what are the most promising enabling areas to be tackled
- Eg data interoperability, security, smart plants, sensor networks, multiscale modeling/simulation, large scale optimization, etc.
- If there is a CBET community, what do you consider to be the leveraged industry problems
- Eg US supply chains (fuels and chemicals), prediction/analysis of toxic drug effects, analysis of commodity processes, zero incident management
- What actions/changes are needed to proceed? What investments are needed?
- Eg community software engineering, data standards, databases
- What are the key CBET CI specifications?
- What are the key take-aways from the workshop?
- What are the architectures that we want? Eg give more input to supercomputer centers
- Tradeoffs of storage, speed, interconnects
- Crystallography – protein structure determination – has migrated from a supercomputing problem to a desktop problem
- Today – what is the problem that requires a supercomputer today and will for the next several years?
- Mistakes – hype over what the next generation will bring, eg gigaflops will give us the cure for cancer … but we need a vision for people to understand the value – perhaps best put as gigaflops that will *help* us get the cure for cancer
- What are the big problems
- Give us a flexible infrastructure that can solve many different problems – too vague?
- Example – sensors, sensor networks, integration with control – large scale problems that are beyond the scope of one individual’s research
- How does a drug affect the whole person – from molecular dynamic simulations at how drug interacts, then gene regulation/network, tissue level, organ level, systems, human, population – Virtual Human
- Different kinds of modeling – molecular simulations; signaling/networks; CFD
- Subproblems under this are grand challenges themselves -- Idiosyncratic tox
- Many potential users could make this tricky to approach
- Particulate processing in pharma industry – manufacturing, storage, delivery – beyond qualitative of a single process
- Supply chain model of the US (and how to accomplish this while managing confidentiality issues) – resource to evaluate impact of disruptions; transportations, energy infrastructure
- Even subsets of this like transportation modeling, sensor infrastructure, etc to predict, manage traffic flows
- Structured information for research – semantic web and literature access, access to disparate databases, etc. Going beyond NLP on Pubmed abstracts or putting bandaids on accessing existing literature/databases. Extending that to include annotation.
- Long term data archive – eg the pharma transition to electronic data submission to the FDA – that data has to live for a very long time – 30-50 years
- Global environment > Language/translation
- Collaborative planning, cross-organization ERP
- Seamless connectivity – fully available, transparent transitions (eg like cell phone connections)
- Process safety – whole infrastructure to support
- Community needs to open up to possibilities
- Eg making safety more than a training issue – what does an infrastructure, systems engineering approach bring to this
- Sharing, creating a community – good examples: geo, weather
- Challenge for communities where modeling is still new, don’t have full buy-in, and no standard approach to modeling – eg in biology where modeling is still new to the wider biology community vs. the bioengineers
|
|
