Everything You Wanted to Know About RadosGW
summary
Overview of RadosGW architecture, current gaps, and concrete fixes: multipart uploads can be 2–3x slower on small objects; versioning plus lifecycle rules free space automatically; bucket sharding moves to automatic bucket partitioning to avoid manual sharding and to keep key ranges balanced; cache consistency bugs get targeted fixes and a proposal for a group-aware cache invalidation; HTTP layer is being replaced by an asynchronous Boost ASIO front-end with a scheduler and QoS, experimental results show much higher connection density and better throughput.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:10What is object storage
- 3:03Cloud object storage
- 3:48Uploading objects
- 5:55Versioning
- 7:54Scalability
- 8:35Object Storage
- 10:22Bucket Index
- 12:17Multistar
- 13:39Replication
- 14:49Search
- 15:47NFS
- 17:42Service Description
- 18:04Metadata Enrichment
- 18:51Cloud Synchronization
- 20:27Scalability scaling
- 25:24Bucket scaling
- 29:23Consistency
related talks
36mUsing NGINX as a CDN for Ceph Radosgw
01h 03mExtreme Performance Architecture
55mThe Future of Rack
21mD4N in RGW
15mBeyond RSS: Towards Intelligent Dynamic Memory Management (Work in Progress)
26mAdopting gRPC for Low Latency, High Bandwidth Scenarios
01h 06mBest practices when accessing Big Data or any other data!
39mKeynote: Lead Developers Town Hall