Lightning Talk: Grasp the Real Worth of Your Microservices by Brewing a Truth Poti...
summary
Algorithmic recommender evaluates real importance by clustering apps by resource requests and comparing median resource and network usage to produce weighted percentile scores that map to priority classes and QoS. Uses N-dimensional K-means clustering, telemetry from OpenTelemetry / Jaeger / custom backends, median-based scoring and weighted ranking, validated against prod traces then applied in pre-prod AB tests. Results show about 70% accuracy in matching human judgments and yield 13% fewer pod evictions and ~8.5% lower capacity.
AI-written summary · about
chapters (6)
- 0:00Microservices priority challenges
- 1:23Static priority limits in Kubernetes
- 3:46Dynamic priority with algorithmic ranking
- 6:37Four-step priority scoring
- 8:44Evaluation results and impact
- 10:08Q&A
related talks
11mLightning Talk: Searching for Serverless Candidates in the Jungle of Microservices
25mKube Intelligence - A Metric Based Insightful Remediation Recommender
06mLightning Talk: Who Are You? A Heuristic Approximation to Service Identification
23mLightning Talk: Rethinking Microservices
36mA practical tale of assessing a Microservices Architecture
09mLightning Talk: Dissecting Service Mesh Overheads
49mDeveloping microservices applications on the Kubernetes platform
13mLightning Talk: Observability-First DevSecOps: Building Resilient Multi-Cloud P...