RoCEv2 Converged Fabric Deployment
Deploy RoCEv2 next to TCP/IP in a data centre: PFC, ECN, DCB, switch configuration and automation.
Advanced30 lessons
written lessons≈ 15 h
High-Performance Networking & Cloud-Native · 30 illustrated lessons (text/HTML)
Outcomes
- Design a lossless Ethernet fabric with PFC and ECN
- Configure switches, NICs and DCB end to end
- Automate and monitor a converged RoCEv2 network
Where we teach this
- AIC-210 · RDMA & AI Cluster Networking live cohort
The lessons
Every lesson in this curriculum, in the order we teach it.
- 1Introduction: RoCEv2 and TCP/IP converged deployment
- 2Networking refresher: TCP/IP and data-centre network architecture
- 3RoCEv2 technology in depth
- 4Core challenges of converged deployment
- 5Network topology design: physical and logical topology principles for converged networks
- 6Switch configuration basics: choosing and setting up RoCEv2-capable switches
- 7Priority and traffic classification: DSCP and PCP marking and priority-based flow control
- 8PFC configuration in practice
- 9ECN configuration in practice: principles, setup and PFC interplay
- 10DCB configuration explained
- 11Host-side network configuration: NIC firmware, driver installation and basic parameter tuning
- 12OS kernel parameter tuning: optimising a system where RDMA and TCP/IP coexist
- 13RoCEv2 communication libraries: installing and configuring libibverbs and librdmacm
- 14Basic connectivity testing
- 15Performance benchmarking tools in practice
- 16Choosing and applying congestion-control strategies: DCQCN, TIMELY and their configuration
- 17Multi-tenancy and network isolation: VLAN and VXLAN in converged networks
- 18Integrating with storage: deploying and configuring NVMe over RoCEv2
- 19Integrating with compute frameworks: MPI over RoCEv2 in HPC
- 20Deploying in virtual environments: enabling RoCEv2 SR-IOV in VMware and KVM
- 21RDMA device plugin and CNI configuration in Kubernetes
- 22Monitoring and alerting: tracking converged-network health with SNMP and telemetry
- 23Introduction: troubleshooting challenges in converged networks
- 24Link-fault troubleshooting: physical links, MTU and auto-negotiation
- 25Troubleshooting protocols and configuration
- 26Diagnosing performance problems
- 27Hardening strategy: access control and best practice in converged networks
- 28High availability and redundancy: link aggregation, multipathing and failover in RoCEv2 networks
- 29Automated deployment in practice: configuration management with Ansible and Python scripts
- 30Course summary and outlook: RoCEv2 and TCP/IP convergence
All-access library
Every curriculum on this page, with the labs and lab sequences, for a year.