For this DVD, Dell Engineering validated the Dell PowerSwitch S5248F-ON running SmartFabric operating system 10 for the top-of-rack (ToR) switches and Dell PowerSwitch S5232F-ON for the aggregation switches.
Component | Operating system build | |
Dell PowerFlex cluster ToRs | 2x Dell PowerSwitch S5248F-ON | 10.5.6.1.100 |
Azure Stack HCI cluster ToRs | 2x Dell PowerSwitch S5248F-ON | 10.5.6.1.100 |
Aggregation switches | 2x Dell PowerSwitch S5232F-ON | 10.5.6.1.100 |
ToR switches
In the Dell lab, one pair of ToR switches was configured to provide network connectivity to the MC nodes in one rack. A separate pair of ToR switches provided connectivity to the PowerFlex nodes in a separate rack. The switches in each pair of ToRs were connected together by a Virtual Link Trunking interconnect (VLTi), offering high availability across the switches. The VLTi needed to have sufficient bandwidth to cater to storage, management, and workload traffic in case any of the server facing ports fail. In the Dell lab, we used two dual-density ports for the VLTi, which broke out to four 100 Gbps switchports. These switchports participated in the VLTi Switch configurations for this solution.
Aggregation switches
Dell Engineering used a pair of PowerSwitch S5232F-ON switches for the aggregation switches. To minimize oversubscription, each pair of ToR switches connected to the aggregation layer with an 800 Gbps VLT port channel. Furthermore, a VLTi was configured using two 100 Gbps physical interfaces between the aggregation switches to ensure N+1 availability and adequate resiliency through the networking design.
The storage traffic between the MC nodes and the PowerFlex nodes were non-routed, Layer 2 network that traversed the ToR and aggregation switches. All ports that were configured for this storage traffic required support for Jumbo Frames for optimal storage performance. Dell does not support a routed, Layer 3 configuration for these storage networks. However, all management networks were routable in the Dell lab.