NVIDIA Helping Drive Project Monterey with VMware to Get Hyperscaler Composability into the Enterprise
Current and Future Enterprise Workloads will Benefit From Significant Improvements in Performance and Agility
Over the past two to three server purchasing cycles, Hyperscalers have increasingly offloaded functionality from the CPU to the SmartNIC to free up valuable CPU cores. As a result, hyperscalers provide significant compute directly in the NIC so the CPU can focus on the workloads and eliminate networking overhead. DPUs (Data Processing Units) can push this even further with advanced security, telemetry, and storage no longer requiring CPU cycles to complete. To date, the enterprise has been left behind on these cloud advances. Project Monterey, a collaborative effort from VMware and hardware providers such as NVIDIA, will help bring composability and advanced CPU offload to the enterprise. VMware is only working with select server and connectivity companies.
What is Project Monterey
Project Monterey is a re-architecture of how and where key VMware components will run. The new architecture will allow NSX and other parts of VMware vSphere to run on the DPU and free up X86 cores on the servers. It is an evolution in enterprise workloads and modern applications to allow them to run similarly to how they would run on an IaaS provider. Project Monterey is just as much about cost and power savings as it is about distributed architectures and strengthening multi-cloud across on-premises, colocation, and IaaS.
NVIDIA’s Role in Advancing Project Monterey
NVIDIA is a leader in server connectivity across the enterprise and Cloud segments with its portfolio of Foundational NICs, SmartNICs, and DPUs. In the Cloud, multiple Hyperscalers rely on NVIDIA for connectivity. If we look at the announced roadmap for NVIDIA’s BlueField-4, we will have integrated GPU and DPU capabilities. NVIDIA offers an open DPU development platform called DOCA for its ecosystem of users and partners to easily access DPU hardware and software libraries. We expect to see new offerings like RDMA and cultivated offerings like L3 switching, deep packet inspection and advanced telemetry. NVIDIA’s cloud and networking expertise will help drive Project Monterey’s performance.
Massive Improvements in Performance
Freeing cores on the x86 and having those offloaded functions run on specialized DPU and CPU cores on the NIC will significantly improve performance. We estimate that three servers with Project Monterey hardware will be equivalent to four servers with today’s foundational NICs. We don’t expect enterprises to remove that fourth server; instead, they will be able to scale in a similar fashion and cost as Hyperscalers can do today. Remember all the talk back when VMware first ramped up that there would be a huge server reduction in data centers because of virtualization? Instead, virtualization led to a massive expansion in the market–each server ran more workloads but the total number of servers still increased. We view Project Monterey through a similar lens with the potential to drive significant growth. Some functions cannot be measured in performance. Security, such as IPSEC, can’t be done in software when using faster network speeds, and the value of avoiding unauthorized data interception is critical, but not measured solely in performance terms. Putting a virtual switch into silicon can’t be easily measured. Nevertheless, we expect significant savings, processing improvement, and power efficiency for every enterprise when they move over.
Strengthening Multi-Cloud at Next-Gen Speeds
Additional cores applied to the revenue-generating workloads rather than infrastructure overhead, will allow enterprises to run at higher speeds in the network. This is especially true as we look at future processor designs from AMD and Intel with even higher core counts. Additionally, we expect server, CPU, and connectivity enhancements to allow enterprises to move from 10/25G at the server to 100G and even to 200G for specific servers supporting AI, virtual reality and technical computing workloads. Such a rollout aligned with Hyperscalers using multiple 100G links will enable the enterprises to further benefit from some of the Cloud hardware advancements.
With VMware Explore (renamed from VMworld) at the end of August, we expect to see updates shortly for Project Monterey. But, more importantly, as customers begin to adopt and see the performance gains they can achieve, we hope enterprises embrace composability and build differently as we approach 2023. This type of architecture will benefit both the enterprise and the supporting vendor ecosystem.