Fine-Grained Isolation for Scalable, Dynamic, Multi-tenant Edge Clouds
summary
EdgeOS implements a microkernel OS that builds featherweight processes and a copy-based memory movement accelerator to give per-client isolation while keeping line-rate networking. It ports DPDK, runs Click, memcached, TLS proxies and NN inference, and offers fast caching/restore of process chains. Measured startup is ~170× faster than Linux processes and ~100,000× faster than containers; latency is ~25× lower than OpenNetVM, throughput reaches ~700k HTTPS rps and sustains hundreds–thousands of isolated instances.
AI-written summary · about
chapters (8)
- 0:00Introduction to edge clouds
- 1:16Requirements and challenges
- 3:50Limitations of current systems
- 5:50Edge OS architecture
- 7:31System components and logic
- 11:17Packet processing workflow
- 12:44Performance evaluation
- 18:50Conclusion
related talks
22mEmpowering Edge Software Stack on Arm with High Performance Container Networking Tech
25mDPDK performance optimization for edge computing scenario
25mKeynote: Network Transformation: Are We Ready for 5G and Edge Clouds?
29mLiving on the Edge with DCN Networking
29mEnabling Fast, Dynamic Network Processing with ClickOS
25mShenango: Achieving High CPU Efficiency for Latency-sensitive Datacenter Workloads
21mNetworking for Distributed Cloud & Edge Computing
25mEyeQ: Practical Network Performance Isolation at the Edge