Many Layers of Availability
summary
Argues that real availability comes from stacked, interoperating layers and gives concrete examples: built a public cloud with kubernetes as a service and leverages autoscaling, anti‑affinity and self‑healing to reach "so many nines" in practice; reports Saga orchestration often breaks in the wild; uses Redis and Kafka replicas, backups and caching plus CDNs to prevent cascades; applies rack‑awareness, multi‑region replication and GSLB anycast for disaster recovery and global failover, all of which proved operationally valuable.
AI-written summary · about
chapters (16)
- 0:00intro
- 0:14introduction to application availability
- 1:04the importance of microservices
- 1:41speaker introduction and background
- 3:10designing highly available applications
- 4:23redundancy and load balancing
- 6:53reducing dependencies and consolidating features
- 9:13asynchronous communication
- 10:45platform and infrastructure
- 11:22cluster architecture and load distribution
- 13:30orchestration and auto scaling
- 16:57data redundancy and caching
- 20:55infrastructure and networking
- 25:38hardware reliability
- 28:16global availability and disaster recovery
- 31:59conclusion and summary
related talks
05mFour Levels of High Availability in Cloud Foundry
41mAn Adventure in Multi-Cloud with Kubernetes
45m⚡A Resilient Application in a Partially Degraded World ⚡
37m5 Levels of High Availability: from Multi-instance to Hybrid Cloud
40mCloud Foundry Availability
30mArchitecting for Active-Active Resiliency in the Cloud
18mScaling APIs to Billions: Extreme-Scale Reliability
49mPatterns for resilient, highly available cloud architectures