“A Crystal Ball to Prioritize Technical Debt” by Adam Tornhill

Information to Prioritize Tech Debt

• Where’s the highest interest rate?
• Does the architecture support the way system evolves?
• Any bottlenecks for inter-team coordination?

File Level Hotspots

• version control holds information on code changes through time
• tools can reveal info about files
• 
  • bigger circles: greater code complexity
  • darker circles: more frequent code change

Prioritize Through Hotspots

• code change frequency much more important than code complexity metrics
  • complexity metrics are pretty bad, we can approximate to lines of code
• hotspots represent where code pays off to be clean, changeable, understandable, etc.
• in practice, hotspots usually are in bad shape, great refactoring candidates

We can have:

• file-level hotspots
• function-level hotspots
• microservice/module/component level (by grouping related files)

Supervise Trends

• 

Normalization of Deviance

• as we accept situations that deviate from standards, we redefine the standards
• e.g. if we get used to a 6000 lines of code file, we tend to accept other large files, and increases in the 6000 lines of code file

Temporal Coupling

• when a change occurs in A, a change in B quickly follows
• A and B are temporily coupled
• we should measure microservice temporal coupling

Avoiding Microservice Shotgun Surgery Antipattern

• changing 1 thing means changing 5 services
• common reasons:
  • sharing code
  • leaky abstractions, depending on implementation
  • same team, several services
• we can use temporal coupling to detect this antipattern
  • this may indicate architecture needs to change

Tools for Analysis

• Code MaaT
• CodeScene
• Evolution Radar
• Moose Platform

Teamwork Implies Process Loss

• process loss: metaphor from mechanical inefficienies like friction and heat loss
• process loss cause:
  • coordination, communication overhead
  • diffusion of responsability

Measure Team Coordination

• we can measure coordination hotspots
• 
  • shading indicates how many people from different teams work on that file/entity
• coordination overhead in such frequent multi-team changes
• worth investigating why such multi-team changes
• feature teams work if aligned with code being changed

Resources

• book Your Code as a Crime Scene
• book Software Design X-Rays
• www.empear.com/blog
• codescene.io