Top N Events
• How is disk info stored?
– Aggregate, device, filesystem
• Missing information from top "N"
– Runtime of app
– Job status info
– Machine status
• How does application info fit into this?
– Is it the target? - part of the event

Top "N" Event Refinement
• To what extent are the events useful on their own?
– Tradeoff between efficiency and customizability
• Package that provides similar (much more extensive) monitoring data:
– SGI's "Co-pilot"
• To-be-discussed: what kind of encoding / transport this would map to
• Use of RDF (DAML+OIL) for formal ontology language to describe events
Targets
• Need to consider targets found on Grid, not other places
• Add partition, queue target types

Naming
• What to use as separator?
– Suggestion: pick something and define an escape character
– (or do we need to define the encoding?)
• Is the event unique (even with the same name)
– Timestamps, values distinguish instances
• Is "event refinement" arbitrary?
– E.g., line between network/bandwidth, bandwith/tcp, tcp/single-stream
– Could just call the whole thing an "event"
Future
• Scope of group?
– Originally: short-term
• Join the mailing list!
• New draft of document
– Explicitly state intended scope
– Reference to GMA use-cases