In a fast-moving field, how should an architect decide what to learn deeply and what to learn shallowly?
Show the full answer Hide the answer
The allocation
Deep in the durable, shallow and current in the volatile.
Durable, worth deep investment: distributed systems fundamentals, data modelling and consistency, queueing and capacity behaviour, security principles, and the economics of infrastructure. These change slowly and explain everything built on top of them, so the investment compounds for decades.
Volatile, worth broad awareness: specific products, frameworks, model families, managed services. Enough to know what exists, what problem each solves and when it applies — with depth acquired on demand when a decision requires it.
Why the fundamentals matter more in a fast-moving field, not less
Every new technology in a domain is a new set of trade-offs against the same underlying constraints. An architect who understands queueing behaviour, amplification arithmetic, consistency models and the economics of storage and bandwidth can evaluate a new offering quickly, because the questions are the same ones.
An architect who knows only the current tools has to learn each new thing from scratch and has no basis for comparison.
The practical habits
- Read the failure literature. Detailed public postmortems are among the most valuable engineering documents available, and they teach failure modes that no product documentation covers.
- Evaluate against your own workload, with a time-boxed spike. Public benchmarks answer questions you do not have, on data that is not yours.
- Learn from the estate you operate, which is the most relevant source available and the most ignored.
- Track the durable, not the announcements. A new capability matters when it changes a trade-off, not when it is released.
- Maintain the ability to re-evaluate, since in a volatile domain the evaluation harness is worth more than any current conclusion.
The trap
Chasing the volatile at the expense of the durable, which produces an architect who is current and cannot reason from first principles — and therefore cannot distinguish a genuine advance from a repackaging of something the industry already learned about.