Lessons learned from the CSPA Proof of Concept at Statistics Canada Robert McLellan

1. Lessons learned from the CSPA Proof of Concept at Statistics Canada Robert McLellan Chief Enterprise Architect April 14, 2014

2. Last year in Paris @ MSIS 2013 • Described our work to refocus Enterprise Architecture at StatCan • Linked ourselves to the work of CSPA • Described our service-oriented approach • Identified changes within the Government of Canada and externally Statistics Canada • Statistique Canada

3. Last year in Paris @ MSIS 2013 • Discussed what is changing about how statistical agencies create production capabilities • Addressed the importance of modular approaches at the business, information, and service component levels • Explored key roles involved in the design, build, assembly, and configuration of solutions Statistics Canada • Statistique Canada

4. CSPA is now a year older … Statistics Canada • Statistique Canada

5. CSPA Proof of Concept & StatCan • Proof of Concept Aims (HLG) • Demonstrate the process of working together • Demonstrate business viability to senior management • Prove the value of the Architecture • Create “user stories” and requirements for the next phase of CSPA-based collaborative modernization • Catalogue • Standards evolution Statistics Canada • Statistique Canada

6. Proof of Concept – HLG Outcomes Statistics Canada • Statistique Canada

7. Lessons learned Statistics Canada • Statistique Canada

8. Lesson #1 - Roles • The CSPA role model provides a clear and useful suite of differentiated roles • The role of each Agency’s catalogue and the need for an international catalogue is clearly identified • The Assembler and Configurer roles will be of increasing importance in the future • Agency feedback from PoC activities demonstrates various levels of maturity in this shift Statistics Canada • Statistique Canada

9. Lesson #2 - Resourcing • CSPA Proof of Concept activities were “Idea Phase” R&D activities • It can be challenging to secure time and effort with the competing focus of “production” priorities • Necessary expertise could be difficult to secure in predictable quantities • Service builder and reusable “software component” experts • Production solution assemblers • Have we “qualified” as “mainstream” ? Transformation Projects Statistics Canada • Statistique Canada

10. Lesson #3 – Legal Issues • Assets may not be “given away” – taxpayer-funded assets • “Right to use”, “Right to modify”, “Right to distribute” may be freely granted • Must currently be administered • Secondary monetization may be viewed as a source of supplementary income • New co-developed solutions are treated differently – shared assets • Plenty of examples in the GPL, OpenSource world that work • Liability for errors ? • E.g. OpenSSL ! Statistics Canada • Statistique Canada

11. Lesson #4 Design & Implementation Models • GSIM is an effective conceptual design model for service definition • DDI 3.1 provided a useful basis as an implementation model • Potential for “overkill” • Gaps in representation • Need for DDI 4 • “Rule languages” are not addressed but are an opportunity • Identified at the Rome sprint Statistics Canada • Statistique Canada

12. Lesson #5 – Getting the data • CSPA identifies two mechanisms • “pass by value” – in the service message • “pass by reference” – a URI pointer • CSPA is agnostic of underlying technology platforms • Capability of underlying platforms may not be ready • Solutions include “data planes”, Data Service Centres, data access layers • Need to transition from shared drives and local team repositories • Accessing the right amount of metadata at the right time is important • Address “captive platform” limitations • Avoid “dogmatic” conversion Statistics Canada • Statistique Canada

13. Lesson #6 – Service Granularity • CSPA doesn’t provide much guidance about what makes a “good” service • Bottom-up approaches run the risk of blindly turning software API’s into services • Ad hoc web services can be created prolifically with great enthusiasm • Business service decomposition can leverage GSBPM but may need more functional decomposition • Methodology architecture (and taxonomies) would help here Statistics Canada • Statistique Canada

14. Lesson #7 – Waiting for Godot* • CSPA is a great start at creating a “concept of use” for our standardization efforts • Statistical Network Business Architecture • Information models from GSIM • Service design, build, assembly, configure • “Tactical wins” • How does it materialize in our business model at each Agency? • Incremental production improvement • Capability development • Strategy linkage * Driven by this * You want to be here * And to some, you are just this * You are here Statistics Canada • Statistique Canada * With credit to Rudi Seljak @ Slovenian Statistical Office

15. So now what do we do… …having successfully sequenced the DNA of Statistical Agencies ? Statistics Canada • Statistique Canada

16. Some possibilities… • Focused treatments • Strategic interventions • Pain-point (disease) remediation • Preventive medicine • Risk counselling • Performance enhancement • Cost optimization • Improved service delivery • Enhanced solution delivery satisfaction Statistics Canada • Statistique Canada

17. EA Next Steps @ StatCan • Leverage Business Architecture model to identify strategic opportunities • Resolve specific integration issues through GSIM-based information exchange standardization • CSPA, GSBPM, and GSIM-based portfolio optimization • Collaborative development and solution sourcing - international • Information Architecture development (Statistical Network) • Enhanced internal service catalogue Statistics Canada • Statistique Canada

18. Business Activity Focal Points International Collaboration & Innovation Research & Portfolio Enhancement Cost Efficiency Portfolio Optimization & Cost Efficiency Statistics Canada • Statistique Canada

19. Project Portfolio Management is key Relevance, Accessibility, Cost Efficiency, Quality, Risk Statistics Canada • Statistique Canada

20. Questions ? Statistics Canada • Statistique Canada

21. Contact information: Robert McLellan, Chief Enterprise Architect Statistics Canada email: robert.mclellan@statcan.gc.ca