Middle of Semester Presentation

1

• Middle of Semester Presentation
• October 22, 2004
• Grant Averett, Adam Gurson, Minh Nguyen, Richard
  OHara

2
Pandora Project Background

• Students are full time employees of Integrity
  Applications Incorporated
• Project objective develop a generic analysis
  framework that could be applied to various
  problem sets

3
Pandora Team Goals

• Create the Process and Management framework
  necessary to successfully complete our project
  and demonstrate to IAI that a more formal process
  can make software development a less risky
  proposition
• Develop a Generic Framework
• Parsers Read incoming data
• Models Expected behavior of a system
• Correlators Compare input data sources and
  expected behavior
• Develop a Demonstration Capability
• Model a satellite system to include parsing of
  incoming data, comparison of planned versus
  actual data, notification of anomalies

4
Current Roles

• Team Lead
• Minh Nguyen
• Customer Liaison
• Grant Averett
• Configuration Management
• Adam Gurson
• SOW Lead
• Richard OHara

5
Current Status

• Proposal Drafts Complete
• Statement of Work Draft Near Complete
• Waiting for clarification from customer to
  identify specific data sources
• Critical requirements have been identified
• Project Lifecycle Defined
• Artifact based lifecycle
• Formal requirements stage with formal
  requirements specification
• Formal design stage with UML artifacts
• Implementation after design completion
• Testing refinement throughout lifecycle

6
Lessons Learned

• Formal agenda and management of meetings lead to
  more productive sessions
• Freeze artifacts 24 hours before meetings
• Allows participants enough time to review
  artifacts more productive discussions

7
Lessons Learned (continued)

• Team Communication Difficulties
• Too much email traffic
• Lack of team coordination
• Risk Mitigation Strategy
• Weekly Team Meetings

8
Current Risks

• 40 hour work week versus demands of studio
  project
• Difficult to align team schedules
• Summer semester of 30 hour studio
• Team progress slow as we are defining roles,
  processes
• Work environment encourages code over process
• Lack of identification of specific data sources
• Prevents completion of SOW, complete definition
  of Use Cases

9
Questions?

• How do you validate requirements that you cannot
  visualize?
• How do you capture quality attributes as atomic
  and testable requirements?
• Will use cases and stereotypes capture functional
  and non-functional requirements?
• How do you plan for changes to your schedule
  because of activities creep?
• How have other teams overcome communication
  problems?

10
Backup Slides
11
The Analysis Triad
12
Parsers What data do we have?

• Convert raw input data into structured data
• Usually one parser for each type of input data
• Store parsed data into structured (but
  uncorrelated) databases
• Could have some primitive internal correlation to
  generate structured data types from multiple raw
  data fields

13
Models What data are we expecting to have?

• Digitized Domain Knowledge
• Can be manually entered by an expert user or
  generated by a correlator from historical data
• Represents known trends or anomalies (i.e. the
  magnetic field over X lat Y lon always causes a
  satellite to temporarily lose communication
  capabilities)
• May need to be updated over time as correlator
  discovers consistent discrepancies between the
  model and parsed data

14
Correlators How do they differ?

• Compare parsed input data against model to
  determine possible anomalies or discrepancies
• Could be used to generate model from scratch from
  historical data
• Could have any number of correlators across
  different data sets, each looking for different
  trends/discrepancies
• Trigger communications (i.e. JMS events) with
  users when anomalies are detected
• Store correlated output data in correlated
  database

15
Possible Architecture
Event Channel
User 1
Receive Anomaly Events
Input Domain Knowledge Into Model
User n
GenerateAnomalyEvents
Query/ViewCorrelated Data
Query/View Raw Data
Models
M
M
Parsed Data
Correlated Data
PD
PD
PD
CD
CD
CD
ReadModel
C
Read Raw Data
P
P
P
Parsers
Correlators
C
Generate Model
Raw Input Data