370 likes | 503 Views
Deployment in the Space Sector. Progress Report 22/09/2008 Timo Latvala. Outline. Current Vision of Deployment Strategy Training Pilot Modelling Tool Feedback Management Issues Conclusions. STRATEGY. Task vs time. Tools and Modelling Dominate in the Beginning. Goals for Year I.
E N D
Deployment in the Space Sector Progress Report 22/09/2008 Timo Latvala
Outline • Current Vision of Deployment Strategy • Training • Pilot Modelling • Tool Feedback • Management Issues • Conclusions
Goals for Year I • Learn B and use of Rodin • Contribution to JD1 (D5): • Achievement of Training Goals • Requirements engineering: does it help us to write better requirements • Modelling approach • Suitability • Complete first stages of pilot • Initial feed back to tools • Measurement goals
Activities for M24 and later unclear KEY QUESTIONS • Going from pilot to enhanced deployment • How to achieve integration with normal SW process? • The pilot deployment is critical for success
Training Goals Have Been Met ACTIVITIES • Blocked Course in Zurich • Internal Training Arranged by Aabo • Mini-pilot modelling Fast learning facilitated by FM experience in RAMS team SW Engineers also have achieved basic understanding
Pilot Model Development More complete BC model Complexity BC Services Mini- Pilot May October January 2009
Pilot Continues... More Complete BC model PUS Services Complete BC model May 2009 January 2009 December 2009 Time
Current Model Captures High-Level Features • All instruments are modelled • Focus on State machine features • Basic TM/TC traffic • FDIR is completely abstract • Instrument HW is not modelled • TC traffic is completey non-deterministic
The most important invariants are relations between state machines
White Paper Defines Modelling Approach • Key Requirements have been identified • First step was to create granular global model • Next refined models of instruments were introduced
Iterate and Compose • Refined instrument models are composed with the global model after every step • Smaller steps makes finding errors easier • Deviates somewhat from the approach in the White Paper
Conventions to Manage Complexity • Naming conventions (global namespace) • Instruments are modelled separately first (lack of team features)
Next Target is November Plenary Meeting • More complete TC/TM • Instrument HW • Fully composed model • Traceability in the model
Feedback to Requirement Documents • Initial instrument modes • Instrument mode transition • Management of house keeping data Actual changes are still TBC
Team Work is not Supported • Model breaks easily breaks if two people work on the same development • Composing models from different developments is difficult
Lack of Modularity is a Show Stopper • Managing complexity is challenging already hard for small models • Modularity in the model is needed NOW • Development of larger models is infeasible without it
Proof Management Still Needs Work • Deafult tactics proove too little, enhanced tactics are too slow, even crash the tool • Work around: enhanced tactics only for certain theorems works some times • Automatic build is a mixed blessing • Training on advanced use could help (best practices)
Partners: Aabo, Cetic, UNew, and Soton are active ACTIVITIES • 1 Training with Aabo • 2 Planning Meetings with Aabo • Soton contributed a pilot model • UNew contributed a pilot model • Cetic has been arranging measurement
Practical form of co-operation still unclear compared to resource allocation... New ways of collaboration are needed!
Complete Next Step of Pilot as Planned • Pilot is in good shape • Modelling proceeding according to schedule Add the planned features for the Plenary Meeting
Bugs and Stability Issues Must Be Adressed • Latest Version of Rodin released 11/3/2008 • Response to bug reports is SLOW More resources must be commited to tool maintenance
Advanced Training Could Increase Modelling Efficiency • Basic things are not a problem • Most issues related to manual proving and capturing Arrange course on advanced topics
New Features are Needed Urgently • Team work is too hard • Complexity management is too hard • We still dislike the type system New features should be implemented earlier!
Partner Involvment Needs To Be Increased • Collaboration is not very active at the moment • Partner contribution is vague and expectations are unclear New forms of collaboration and explicit plans should be defined
Increase SW Engineering Involvment • SW engineering team not very involved after initial training period • We will pay for it later unless we act Bi-weekly meetings with SW team on modelling
Decided actions • SSF to submit buggy models to BSCW to show crash problems • SSF to submit models which show what issues should be covered at advanced course • SSF to submit pilot model for comments and then suggest telecon date. • SSF to start co-ordinating meetings with SW team on pilot model