The underlying physical network infrastructure must support high network throughput and low network latencies on the data path of the data plane intensive CNFs.
Consider the following best practices when you set up a high performing physical network infrastructure:
Choice of NIC: Before selecting a NIC type, you must have a thorough knowledge of the capabilities and performance of that NIC. NIC throughput, host PCIe speeds, and CPU offload impact the performance of the data plane intensive CNFs. The secondary network (either VMXNET3 with Enhanced Data Path capabilities or SR-IOV) that you choose is also an important factor as both features must be supported by the NIC vendor.
Important:In case of SR-IOV, the CNF vendor must support the chosen NIC as this type of acceleration bypassed the ESXi hypervisor. For a more comprehensive analysis of the pros and cons of Enhanced Data Path and SR-IOV, see the Secondary Network for Data Plane Pod section.
Server Homogeneity: The server model and configurations used for data plane CNFs must be identical. For example, the same NIC and NIC firmware must be used in all hosts used for data plane CNF.
NUMA Node Design: If multi-socket servers are used for data plane intensive CNFs, design all the Non-Uniform Memory Access (NUMA) nodes identically. For example, if the data plane in a single NUMA uses two NICs, other NUMA nodes must also use two NICs for the data plane applications. This guideline applies to other components that have NUMA locality such as memory and CPU. It does not apply when using single-socket servers.
Data Plane Traffic Path: The speed of any network is determined by its slowest link. To drive the maximum performance out of VMware Telco Cloud Platform that runs the data plane intensive CNFs, the NICs, Top-of-Rack (ToR) switches, and End-of-Row (EoR) devices must support high data plane performance. Ensure that you reserve the NIC ports and ToR switch ports for the data plane traffic and isolate them from the control and management plane traffic.
Data Center Fabric Design: The choice of a data center fabric is crucial to provide a high-performance, scalable, and efficient network infrastructure to interconnect computing resources. Choices such as Layer 2 versus Layer 3 Spine/Leaf Architecture, teaming and link aggregation policies and strategies for Load Balancing and North-South communications must be considered carefully.
For information about the Data Center Fabric design alternatives, see the Network Architecture section in the VMware Telco Cloud Reference Architecture Guide 3.0.
