Technology practitioner | CTO @PrizsmTech | Founding Director @Fern_ICT | Architecture, AI, cyber resilience & information infrastructure
Chalford, Gloucestershire, UK
Joined March 2012
- Tweets2K
- Following1.8K
- Followers2K
- Likes1.2K
The Celebrity Traitors is essentially a distributed trust problem. Nobody has complete information, everyone has a different view, and consensus doesn't guarantee correctness. Sound familiar?
67
Distributed systems fail in ways that are difficult to reproduce locally. Google's SRE guidance on monitoring is worth a read: sre.google/sre-book/monitori… #SRE
2
66
If you can't reliably rebuild the software you tested, how confident can you be that you're deploying the same thing?
1
60
A useful incident review asks which conditions made the failure possible, not simply who made the last mistake.
1
62
An average can hide a terrible user experience. For latency, look at percentiles (especially p95 and p99), not just the mean.
1
68
If a consumer can't keep up with a producer, adding a queue only delays the problem. Define what happens when the queue fills.
1
62
Give a service only the permissions it needs, and only for as long as it needs them. A compromised component shouldn't inherit your whole system.
1
58
A good API makes valid operations easy and invalid states difficult to express. Convenience matters less than clarity at the boundaries.
1
58
Don't just test that a system works. Deliberately remove a dependency and observe what fails, what degrades and what recovers.
2
129
A safe database migration should consider both the old and new application versions running at the same time. Deployments aren't instantaneous.
1
3
116
Every remote call needs a timeout. Otherwise, a failed dependency can turn waiting into resource exhaustion.
2
78
If an alert tells you something is wrong but not what matters, you haven't finished designing the alert.
2
76
Tests don't prove software is correct. They give you evidence that specific assumptions still hold true.
2
64
A system that scales by adding resources isn't necessarily scalable. The real question is what eventually becomes the bottleneck.
3
76
Good documentation doesn't describe everything. It tells the next engineer what they need to know to make a safe decision.
2
112
Every dependency is also a failure mode. The important question is whether you've designed for its absence.
1
87
Configuration is code with a different deployment mechanism. Treat it with the same discipline.
1
73