As the invitation from Intel, our speaker sharing the main architecture design of ChubaoFS and the key features.
- High performance: Optimized for both large and small files.
- Multi-Tenancy: Share the whole storage and resource utilization ratio.
- General- Purpose Storage Engine: Support both sequential and random read/write.
- Highly Scalable: Individual meta and data node clusters. Scales with meta and data partitions increase.
- POSIX filesystem APIs: Complaint whenever possible, and compromise between semantics and performance.
Meanwhile ChubaoFS user scenarios were shared and increasing number of people interested in these.
- Underlying storage for in-memory DB: Preserve snapshots.
- Cloud storage for AI platform Training service: Decouple computing from storage.
- Latency-sensitive services: Used for storing Nginx logs.
- Elasticsearch: Takes care of replications and high availability.
- Service MySQL: Running MySQL history tables on ChubaoFS directly since most of the requests are read.