Good Enough Architecture

Stefan Tilkov takes a look at some of the ways you can determine whether the development efforts you’re undertaking suffer from too much or too little focus on architecture. You’ll examine a number of real-world examples that are intended to inspire either admiration or terror and try to find some recipes of how you can get more of the former and less of the latter in your own projects.

Transcript

1. “Good Enough” Architecture Stefan Tilkov stefan.tilkov@innoq.com @stilkov O’Reilly SACon Berlin, 2019 "Tegel Airport TXL Berlin May 2012 - 19" by andynash is licensed under CC BY-SA 2.0
2. @stilkov (Software) Architecture Definitions Fundamental concepts or properties of a system in its environment embodied in its elements, relationships, and in the principles of its design and evolution (ISO 42010) Whatever the architect considers important enough to merit their attention Architecture represents the significant design decisions that shape a system, where significant is measured by cost of change (Grady Booch)
3. @stilkov Architecture is not an upfront activity performed by somebody in charge of telling everyone else what to do
4. @stilkov Architecture is a property of a system, not a description of its intended design
5. Pick the best car:
6. @stilkov Quality https://iso25000.com/index.php/en/iso-25000-standards/iso-25010
7. Scaling Dimensions Logic Load written in a day depends on German tax law a dozen users half of the planet Netflix Twitter Insurance Policy Management Facebook Amazon Random CMS
8. @stilkov There is no “good” or “bad” architecture without context; architecture needs to take specific quality attributes into account
9. Cases
10. @stilkov Context: • … Observation(s): • … Lesson(s) learned: • …
11. #1: Non-extensible Extensibility
12. @stilkov Context • E-Commerce (retail) provider • Global customer base • Catalog/CMS/Shop/Fulfillment • Multi-tenant • Highly customizable
13. @stilkov Large customers
 (“strategic”) Small customers
 (“long tail”) Specific General High Low (Acceptable)
 Cost Customization
 (Need) The Solution
14. @stilkov If your design attempts to satisfy everyone, you’ll likely end up satisfying no one
15. @stilkov Highly specific code is often preferable to sophisticated configuration
16. #2: Perilously Fine Granularity
17. @stilkov Context • Large-scale B2B food retailer • New company-wide shop and logistics system • >200 developers
18. @stilkov Team 1 Team 3 Team 2
19. @stilkov Order Service Support Fulfillment Billing Checkout
20. Why would you cut up your system into tiny, distributed, hard-to- manage fragments?
21. @stilkov Everybody wants to be Netflix, but nobody is
22. @stilkov
23. @stilkov …
24. @stilkov Order Service Support Fulfillment Billing Checkout
25. @stilkov Support Fulfillment Billing Checkout Order Service
26. @stilkov Lessons learned • Small is not always beautiful • Refactoring within team boundaries much easier than globally • Ignore organizational parameters at your own risk
27. #3: Your system WILL be dynamic
28. @stilkov Context • Large-scale insurance system • Model-driven development • > 100 developers • 2 Releases/year
29. @stilkov Modeling Business Concept 0 3 6 4 5 2 1 Technical Concept Implementation Module Test Integration Test Rollout Vision What if you miss your slot? Modeling
30. @stilkov Policy Holder Clause - id - value - dueDate - name - address - status - text - validity - taxClass * * regionCode Model Name New Name (Meaning) Description Release Introduced taxClass regionCode … 10.3 …
31. @stilkov Lessons learned • Centralized responsibility hurts • Faced with too much rigidity, a way around the rules will be found • Just because you’re used to it doesn’t mean it’s acceptable
32. #4: Horizontal Conway Split
33. @stilkov Context • Unified communication platform • Straight-forward business logic • High scaling requirement • 1-2 teams/10-20 developers
34. @stilkov Group 1 Motto: »Java is a legacy programming language used in the last century« Group 2 Motto: »Obviously, you can’t build correct programs with JavaScript«
35. @stilkov Group 1 Motto: »Java is a legacy programming language used in the last century« Group 2 Motto: »Obviously, you can’t build correct programs with JavaScript« HTML/CSS/JavaScript Frontend Java Backend JSON API
36. @stilkov Lessons learned • Every meaningful feature development required frontend and backend parts, and thus co-operation between teams • When faced with unresponsive backend teams, frontend teams quickly become secret full-stack teams
37. @stilkov Mel Conway (@conways_law, follow him) is right most of the time
38. @stilkov There are no limits to architectural complexity people will accept to stay among their own
39. #5: System of Systems
40. @stilkov Context • E-Commerce/Online shop (Retail) • 100-120 developers • ~10 self-contained teams
41. @stilkov number of developers strength of decoupling methods modules components μservices systems
42. @stilkov From a layered system … System Logic Data UI Module Module Module
43. @stilkov … to a system of systems System System System Logic Data UI Logic Data UI Logic Data UI
44. @stilkov In-page Cross-page JavaScript method calls Links & redirects Shared abstractions & frameworks Micro-architecture Common language runtime HTTP HTML 5 JS platform Standard Browser
45. @stilkov Extremely loose coupling requires very few rules, but they need to be enforced strictly
46. #6: Free-style Architecture
47. @stilkov Context • E-Commerce/Online shop (Retail) • 100-120 developers • ~10 self-contained teams
48. @stilkov But … • Lack of standardization led to inefficient UI integration at runtime • Vast differences in API style, formats, documentation created needless extra work • Despite no centralised frontend, a central frontend team created a new bottle neck
49. @stilkov You cannot decide to not have an architecture; if you don’t actively create it, be prepared to deal with the one that emerges
50. @stilkov There’s a fine line between diversity (that adds value) and chaos (that doesn’t)
51. #7: Cancerous Growth
52. @stilkov Context • Financial services provider with independent brokers as clients • ~30 developers • 20 years of company history
53. @stilkov Oracle DB Oracle Forms App
54. @stilkov Oracle DB Java/JSP Web App Lots of PL/SQL
55. @stilkov Oracle DB Java/JSP Web App Lots of PL/SQL .NET Web Service .NET Web Service .NET Web Service .NET Web Service
56. @stilkov Oracle DB Java/JSP Web App Lots of PL/SQL .NET Web Service .NET Web Service .NET Web Service .NET Web Service Oracle DB Java/JSP Web App Lots of PL/SQL .NET Web Service .NET Web Service .NET Web Service .NET Web Service Company A Company B
57. @stilkov Java/JSP Web App Lots of PL/SQL .NET Web Service .NET Web Service .NET Web Service .NET Web Service Oracle DB Java/JSP Web App Lots of PL/SQL .NET Web Service .NET Web Service .NET Web Service .NET Web Service Company A Company B
58. @stilkov Java/JSP Web App .NET Web Service .NET Web Service .NET Web Service .NET Web Service Oracle DB Java/JSP Web App .NET Web Service .NET Web Service .NET Web Service .NET Web Service Company A Company B Mongo Couch/Pouch Mongo MySQL
59. @stilkov Java/JSP Web App .NET Web Service .NET Web Service .NET Web Service .NET Web Service Oracle DB Java/JSP Web App .NET Web Service .NET Web Service .NET Web Service .NET Web Service Company A Company B Mongo Couch/Pouch Mongo MySQL C++ Encryption Lib
60. @stilkov Lessons learned • Successful systems often end up the worst architecture • Unmanaged evolution will lead to complete chaos • Don’t be afraid of some light architectural governance
61. #8: Improve with Less Intelligence
62. @stilkov Context • Bank with multiple CotS systems • Highly proprietary integration solution phased out by vendor • Project launched to replace commercial product with open source solution
63. @stilkov Magical Integration Broker Custom App CotS DB External Partner Other Group Company Parent Company External Partner Custom App Custom App CotS
64. @stilkov Magical Integration Broker Custom App CotS DB External Partner Other Group Company Parent Company External Partner Custom App Custom App CotS Magical Integration Broker Routing Conversion Transformation Transcoding Transport Error handling Business logic …
65. @stilkov Custom App CotS DB External Partner Other Group Company Parent Company External Partner Custom App Custom App CotS Pub/Sub Messaging
66. @stilkov Custom App CotS DB External Partner Other Group Company Parent Company External Partner Custom App Custom App CotS Pub/Sub Messaging Pub Sub Messaging Pub/Sub routing Transport Error handling Adapter (Docker Container) Conversion Transformation Transcoding Error handling Business logic
67. @stilkov Lessons learned • Smart endpoints, dumb pipes (cf. Jim Webber) • Value of specific over generic solutions • Micro architecture with blueprints
68. Takeaways
69. @stilkov 1. Don’t be afraid of architecture
70. @stilkov 2. Choose the simplest thing that will work
71. @stilkov 3. Create evolvable structures
72. @stilkov 4. Manage your system’s architectural evolution
73. @stilkov 5. Don’t build road blocks – create value and get out of the way
74. @stilkov Stefan Tilkov @stilkov stefan.tilkov@innoq.com Phone: +49 170 471 2625 innoQ Deutschland GmbH Krischerstr. 100 40789 Monheim am Rhein Germany Phone: +49 2173 3366-0 innoQ Schweiz GmbH Gewerbestr. 11 CH-6330 Cham Switzerland Phone: +41 41 743 0116 www.innoq.com Ohlauer Straße 43 10999 Berlin Germany Phone: +49 2173 3366-0 Ludwigstr. 180E 63067 Offenbach Germany Phone: +49 2173 3366-0 Kreuzstraße 16 80331 München Germany Phone: +49 2173 3366-0 @stilkov That’s all I have. Thanks for listening!
75. Rate today ’s session Session page on conference website O’Reilly Events App
76. @stilkov www.innoq.com OFFICES Monheim Berlin Offenbach Munich Hamburg Zurich FACTS ~150 employees Privately owned Vendor-independent SERVICES Strategy & technology consulting Digital business models Software architecture & development Digital platforms & infrastructures Knowledge transfer, coaching & trainings CLIENTS Finance Telecommunications Logistics E-commerce Fortune 500 SMBs Startups
77. @stilkov Growing architectural maturity means less guidance and rules are needed
78. @stilkov The more experienced you are at (active and passive) architectural governance, the less you can do of it