Information Systems Architecture skeletons: Overview Lessons-learned from largescale software-intensive system development

Transcription

1 Information Systems Architecture skeletons: Overview Lessons-learned from largescale software-intensive system development March 9, March 2010 Neil Siegel Sector Vice-President & Chief Engineer Member, National Academy of Engineering

2 Agenda What kind of systems are we talking about? What are the typical symptoms of serious development problems for these types of systems? What are the underlying root causes? One approach to a solution Findings 2

3 Example system: Battlefield digitization (aka Blue Force Tracking ) The use of information technology to increase the effectiveness of the U.S. land combat team Where am I? Where are my buddies? My allies? Where is the enemy? Where are the dangerous areas? What is the route over which I am to move? On the move... all the time Every Army platform... wheels, Battalion tracks, / rotary wing, dismount Brigade AFATDS ASAS FAAD CSSCS C 2 I ASIP SINCGARS EPLRS INC Handset computer MCS R INC EPLRS Brigade / The world s most flexible battalion / and adaptive mobile company wireless network network Significant intellectual property Platoon computer Handset INC ASIP SINCGARS Platoon network ASIP SINCGARS EPLRS ASIP SINCGARS EPLRS INC INC Handset computer Handset computer 3

4 Other examples US U.S. Forward-Area Air Defense System Counter-RAM (rockets, artillery, mortar) system Tactical High-Energy Laser Command Center Processing and Display Systems Various large radar systems Logistics automation systems Air traffic control systems Many others 4

5 What are the typical symptoms of development problems for these types of systems? Performance &/or capacity is substantially below that required Response-time is substantially worse than that required Availability &/or mean-time-between-failure is substantially below that required External interfaces do not work as specified 5 The key lessons-learned about the root causes underlying the above: It s the unplanned dynamic behavior. If you don t design with these risks in mind, you might not see the problem until very late in the program.

6 Programmatic needs Need to start analyzing and debugging dynamic behavior much earlier in the life-cycle Enable implementation and exercising of actual external interfaces much earlier in the life-cycle Facilitate prototyping and benchmarking, thereby providing more visibility into progress Provide technical tools that allow us to meet software schedules and costs Put something into the design that allows us to consider the performance, capacity, response-time, and availability goals And in general, lower the risk that bad things don t show up until systems integration 6

7 One approach: the system architecture skeleton (SAS) External Clients L Band Lower TI Upper TI PASS LAN Xcvr Inc PLGR Data Store Data Catalog Abrams SM MFCS Stryker Bradley Protocol Stacks Agent Router Comm Mgr Message Agents Device Agents SA Forwarders NM HM CM Public APIs / Surrogates Future NGC Apps Auto Msg Mgr C2 Msg Mgr SA Mgr Own Blue Red Georef Air GUI Processing Map Server App1 App 2 App 3 Infra- Structure Core Asset Application Core Asset Legend GUI Core Asset Third Party Application App 4 7 Third Party Applications Build and exercise the system architecture skeleton: Defines and implements all processing paths through the system. Mechanize the threads through some reasonably deterministic tool, allowing one to prevent unplanned dynamic behavior. Populate the SAS at first with stubs, models, and prototypes Integration consists largely of replacing these with actual products

8 Findings Seems to help Many successes, over many years, in many different customer contexts Not strongly dependent on any particular implementation technology Core ideas have been re-implemented in several successive generations of technology Appears to decrease integration risk by: Increasing the time available for performing integration Allows the implementation, integration, and debugging of the system control flows largely separate from the functionality 8

9 9 Questions?