COSMOSDataCollectionMeeting20061219

Agenda

• Status updates and agenda changes (optional)
• Refine release 1.0 deliverables and plan

Attendees

• Don Ebright
• Craig Thomas
• Hubert Leung
• Sheldon Lee-Loy
• Mark Weitzel

Discussion

The December 26 meeting is canceled and the next meeting will be January 2, 2007

The group discussed the need to obtain a better understanding of TPTP architecture to facilitate discussion of common deliverables between our projects. Marius is our subject matter expert on TPTP data collection, but we agreed that the group needs to become more familiar with TPTP.

We discussed database schemas without arriving at a conclusion. We can't reuse the existing TPTP persistence mechanism because it is based on EMF and we have decided not to employ EMF at this time. We noted that the schema must be flexible enough to persist a wide variety of data and must perform well. The discussion was tabled until we have defined our reporting requirements so they can serve to point out the requirements. The most promising approach for the near term appears to be persistence of plain Java objects.

The discussion of data sources centered around short term use of the TPTP RAC and longer term supporting JMX and WSDM. Performance issues may force an approach that delivers large volumes of data through a dedicated connection in addition to the lower performing but higher function of some of these interfaces.

We discussed the data collection communication and normalization layer. The RAC was proposed as an initial data collection technique. We clearly want to support JMX and WSDM in the future. The question of whether it will be necessary to persist the SML was discussed without reaching a conclusion.

We discussed the topology of the monitored environment and whether we need to persist topology information. We agreed that this is desirable but probably out of scope for the near term deliverables due to complexity. For now, we expect that the data that will be collected will contain an identifier that uniquely identifies the configuration item being monitored.

We discussed Sheldon's mockup of a COSMOS uptime report and the need for a mockup of a report that shows statistical (time series) data.

Action Items

• All become more familiar with TPTP 4.3
• Don post a request for a F2F to the TPTP PMC mailing list
• Don email future meeting announcements to all prior attendees
• Don post a draft of a high level architecture and initial requirements to the wiki
• Craig create a mockup of a report containing statistical data
• Mark locate Marius's TPTP data collection bugzilla