eBay Geo-Distributed Database on Kubernetes [A]
summary
eBay runs a geo-distributed document DB on Kubernetes with thousands of storage engines, using a local volume plugin that exposes bare-metal SSDs as PV/PVCs and a scheduler-side PV/PVC bonding to enforce volume-level anti-affinity. Performance class offers ~50K IOPS and up to 500 MB/s; standard class ~300 IOPS/300 MB/s. Local SSD 4KB writes measure ~200 ns vs ~1,800 ns on iSCSI. Cross-cluster pods get global IPs for replication, and snapshots send deltas to remote volumes via a workflow-driven control layer.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:25Agenda
- 0:59eBay Geo Distributed Database
- 2:05Overview
- 6:29Challenges
- 9:40Solutions
- 15:40WISB Example
- 18:13Storage Classes
- 19:11Local Volume
- 28:13Network Volume
- 31:24Backup Solution
- 32:59Future Work
related talks
32mRunning eBay’s High-Performance Workloads with Kubernetes
34meBay Search On K8s
50mRunning Databases Efficiently at Scale - Case Study of Flipka...
38mModel and Operate Datacenter by Kubernetes at eBay
28mBuilding and Managing Kubernetes with Kubernetes
28mBuilding and Managing Kubernetes with Kubernetes
19mBuild Global-Scale Apps Using Globally Distributed Autonomous Databases
25mContainer Attached Storage: Open Source Extensions: Kubernetes + Data
