Enterprise software
Six-fold latency reduction by decomposition
A legacy monolithic pricing engine taken apart into twelve services, cutting median latency six-fold and maintenance cost by well over half.
- median response time
- 6×
- median latency improvement
- 60%
- reduction in system maintenance cost
- 12
- independently deployable services
A pricing engine had accreted for a decade. Every quote ran the whole monolith. A twelve-second median response was accepted as the cost of a complicated domain.
Diagnosis
The real problem was coupling. The slow path touched subsystems it had no business touching, because no boundary stopped it. Latency was structural, so profiling could not fix it. Profiling finds hot code, not bad seams.
Approach
We decomposed along domain boundaries instead of technical layers:
- Event storming the pricing domain to find the real invariants before any code moved.
- Strangler-fig migration. Each extracted service took live traffic behind the existing interface, so there was no big-bang cutover.
- Twelve services, each owning its data and deployable without a coordinated release.
Result
Median response time went from 12 seconds to 2, a six-fold improvement. Overall maintenance cost fell 60%, and that saving is what funded the work.