The book I’m currently reading is Designing Data Intensive Applications by Martin Kleppmann. During the first read, I found concise fundamentals we need to concern most when dealing with a high quality software system. There’re 3 fundamentals as follow:

FundamentalDefinition
ReliabilityThe system should work properly even when the things go wrong (faulty)
ScalabilityThe system should be able to handle the growth (data volume, traffic, or complexity)
MaintainabilityThe system should easily adaptable enough during maintenance (bugs, failures, and new features)

Reliability⌗

Faulty can come up from several places:

  • Hardware Faults: eg. Hard disk has MTTF (mean time to failure) of about 10 to 50 years.
  • Software Faults: eg. Uncaught bugs in Linux kernel due to leap second on June 20, 2012.
  • Human Errors: eg. Configuration error by operators leads the most system outage.

Scalability⌗

Data growth is equal to increasing load. There’re 2 things need to describe:

  • Load: current load existing on the system. Some parameters:
    • request per second
    • response time
    • read write ration on DB
    • number of concurrent user
    • hit rate on cache
  • Performance: metrics that used to define how good the system handle load. It can be used as SLA / SLO. Some metrics:
    • avg / mean
    • median: p50
    • percentile: p95, p99, p999

Maintainability⌗

The goal is to minimize pain during maintenance. Three design principles need to follow:

  • Operability: make it easy for teams to keep the system run smoothly.
  • Simplicity: make it easy for new engineer to understand the system.
  • Evolvability: make it easy for engineer to extend the system.

References

  1. Designing Data Intensive Applications by Martin Kleppmann