CyberStar: Simple, Elastic and Cost-Effective Network Functions Management in...
summary
Cloud network functions run on ECS (containers/VMs) with a three‑tier SC/PP/fabric architecture that pools resources, shares stateless packet‑processing units across tenants, and assigns stateful service‑compute groups on demand. A fabric master caches flows and hides PP scaling; state partitioning and coordinated migration let service groups split/merge during scale‑out/scale‑in. Deployed for years, CyberStar reports 100 Gbps (6.7×) throughput, millions of connections within one second (20×) scaling, and a 100M‑connection test at 35k conn/s for 60 minutes.
AI-written summary · about
chapters (11)
- 0:00CyberStar introduction
- 1:49Traditional NFM challenges
- 3:13Scalable NF requirements
- 5:01Stateless NF design problems
- 6:51SAR architecture components
- 9:03Service computing and state
- 11:12State partitioning for scale
- 13:20Fabric Master's scaling role
- 14:29Fabric mesh for tenants
- 15:10Autoscaling results
- 16:32Q&A
related talks
18mCStar Gateway: Augmenting Public Cloud Infrastructure for Heterogeneous Network...
25mStardust: Divide and Conquer in the Data Center Network
16mPreview: Cloud Computing
35mThe powerful simplicity of Amazon ECS
20mSimplifying cost-optimized cyber response & recovery at massive scale
16mA novel Cloud High Performance Computing infrastructure for user
32mIf your future is on the edge, StarlingX wants you
19mSeamless networking across the edge, the private cloud, and AWS