Speaker - Designing Systems That Contain Failure - CS Week Perú 2026

On August 13, 2026, I spoke at CS Week Perú 2026, an IEEE Computer Society student chapter event across Peru. My session, "Isolation and Trust Boundaries in Production: Designing Systems That Contain Failure," focused on limiting the impact of failures in production systems. The key idea was not to prevent every failure, but to control its blast radius.

Production systems inevitably fail, and failures can overlap, causing crashes, exhaustion of resources, compromised credentials, and unavailable dependencies. Reliability engineering doesn't assume these issues won't occur; instead, it decides what can be affected when they do.

The talk explored three dimensions of trust boundaries: context, identity, and execution. Context involves logical isolation, ensuring request-specific information doesn't contaminate shared processes. Identity addresses cryptographic isolation, using tenant-specific cryptographic material to reduce the blast radius of compromised keys. Execution boundaries tackle physical or process isolation, delegating resource-intensive tasks to independent worker processes or containers with resource limits.

Invariant properties that must never be violated were crucial. These properties must survive production testing, not just be verified in isolated unit tests. The integration suite for the session's system included 61 checks covering various production scenarios, resulting in 61/61 PASSED. However, this doesn't guarantee the system is mathematically proven to be perfect; it simply demonstrates that the defined properties remained true under the evaluated conditions.