Home > Servers > PowerEdge and VMware > White Papers > DPUs in the new vSphere 8.0 and 16th Generation Dell PowerEdge Servers > So, what happens if you replace the NIC with a Data Processing Unit-DPU?
When a NIC is replaced with DPUs, networking, security, and storage functions can now offload onto the DPUs directly, which accelerates their performance while reducing the host resources required to run them to near zero. And when fewer host resources are consumed by infrastructure services, more resources are available for production workloads. In this scenario, except for Host/VMM management and compute virtualization features running on the server, the network, security, and storage functions run directly on the DPU.
The DPU devices available today contain embedded ARM-CPUs, memory, a high-performance network interface card, and programmable data acceleration engines. A trimmed ESXi 8.0 hypervisor is installed directly on the DPU, and various activities are optimized for I/O, packet offloads, external management functions can be directly operated from the DPU. DPU devices are like specific function resource servers with multiple general-purpose compute cores, hence the host server CPU no longer has to process the networking and storage activities. DPUs do not require separate management. All operations that are used to manage the life cycle of the software and firmware of the hosts are used to manage the software and firmware life cycle of DPU devices in the DPU-based environment. To leverage the accelerated performance advantages the vSphere Distributed Services Engine offers, the host servers must include supported DPU device. DPU devices cannot be separately purchased and then installed onto a legacy or non-DPU servers.