Safe Client Behaviour
summary
Shows concrete fixes: mobile apps that poll every five minutes should add jitter (example: uniform between 2.5–7.5 minutes) to avoid synchronized spikes, and avoid truncating randomized times which accidentally resynchronizes clients. Recommend capped retries (e.g. three attempts or max 5-minute delay), use exponential back-off with jitter, and classify errors so only network-timeouts are retried. Move control server-side where possible (use Retry-After header, retry budgets in microservices, and server-side prioritization) to limit cascading overloads.
AI-written summary · about
chapters (13)
- 0:00The problem of client behavior
- 1:36Scale and challenges
- 4:20What makes a request unsafe
- 6:10Implementing jitter
- 10:32Optimizing synchronization
- 13:50Rules for safe retries
- 16:50Exponential back-off
- 21:33Retry conditions
- 22:46Server side control
- 25:58Request prioritization
- 27:20Microservices strategies
- 29:14Adaptive throttling
- 30:33Key takeaways
related talks
15mBoosting System Resilience: Exponential Backoff
32mHow does your application behaves when everything goes wrong?
42mDefinition Of Insanity
35mDesign patterns for application resiliency
01h 01mLittle things and big things: lessons for building large systems.
34mDistributed Systems - What Can Go Wrong and How to Get It Right
29mResilient by Design
42mDesigning for failure