Unlock 2x DPU capabilities with vSphere 8.0U3
![](https://images.core.vmware.com/sites/default/files/vDSE_A_S.png)
Starting with vSphere 8.0, Data Processing Units (DPUs) or SmartNICs were introduced as supported by vSphere and later on by VMware Cloud Foundation (VCF). The support of such devices brought increase performance, enhanced security and greater VM consolidation ratios to vSphere environments. However, at the time only one DPU was being supported due to complex requirements for failover, etc.
We heard loud and clear for the need of dual DPU support, and that is exactly what vSphere 8.0U3 is delivery. I’m happy to announce that starting with vSphere 8.0U3, dual DPUs are now supported, delivering this solution in conjunction with NVIDIA DPUs and Lenovo servers.
Data Processing Units are performance NICs with a programmable processors and built-in accelerators, allowing for the offload of services when utilized with VMware Cloud Foundation. Since these DPUs act as really smart NICs, the need for redundancy and high availability was a must for many customers. Now, with vSphere 8.0U3, customers can have two DPUs per ESXi host acting in two different modes. Customers can choose to deploy DPUs in an active/standby configuration or utilize two DPUs per ESXi host for independent workloads allowing for greater capabilities. Let’s Explore these two configurations.
High Availability DPU Configuration
This configuration utilizes two DPUs in Active/Standby high availability and provides redundancy in the event one of the DPUs should fail. In this configuration, both DPUs are assigned the same NSX backed vSphere Distributed Switch and protects against PSOD and/or DPU uplink failure. DPU failover will occur automatically in the event of a failure, similar to the way NICs failover in current vSphere offerings.
Increased Network Offload Capacity Configuration
This configuration utilizes two DPUs as independent units. Each DPU is directly attached to its own vSphere Distributed Switch. Similar to the way single DPU configuration function, there is not failover as each DPU is the only interface presented to a vDS. However, having two separate DPUs working independently offers and increased offload capacity per ESXi host allowing for additional available throughput for workloads.
Dual DPU Lifecycle Management
Since we are running ESXi on the host and on each of the DPUs within the host, we are talking about 3 ESXi instances per host. One of the key advantages of leveraging DPUs with VMware technologies is the ability to utilize vSphere Lifecycle Management (vLCM) to update/upgrade all three instances of ESXi within the same operation. There are no additional operations needed to update/upgrade ESXi or keep all of the versions at the same release level. This is an industry first.
Additional Resources:
https://core.vmware.com/blog/vsphere-dpus-now-available-vsphere-8
https://core.vmware.com/blog/rise-dpus-infrastructure