Broadcom: Lenovo launches dedicated Vmware servers for the on-premises cloud

Lenovo has used Broadcom's in-house VMware Explore show to unveil a server series optimized for VMware's flagship VMware Cloud Foundation (VCF) product line. In the development of the new Lenovo ThinkAgile VX series with VMware Cloud Foundation, both manufacturers have worked together as part of a co-engineering process and optimized the servers in a number of ways to make it easy to set up an on-premises VMware cloud. Customers using the ThinkAgile VX650-based offering will receive the servers pre-configured for VMware VCF directly from the factory. Lenovo's Professional Service department provides support for installation in the customer's own data center or at the hoster. Lenovo's Professional Service and Premier Support is the central point of contact for all support questions.

Lenovo installs either the latest fifth-generation Intel Xeon processors or AMD's fourth-generation Epyc processors in the servers. The servers support the latest DDR5 memory technology and offer PCIe 5.0 interfaces for the installation of GPUs for AI and VDI applications. On the storage side, VMware's vSAN storage service is already configured in VCF; the servers allow the operation of VMware's new vSAN Express Storage Architecture (ESA) as well as vSAN's original operating mode Original Storage Architecture in all-flash and hybrid combinations (flash and hard disks). Depending on the configuration, customers can use the vSAN functions already licensed by VCF such as de-duplication, compression and encryption for their storage services.

Operation with one DPU – or two

If customers wish to use the Data Processing Units (DPUs) supported since version 8 of VMware's hypervisor vSphere, this is possible in certain server types of Lenovo's ThinkAgile VX650 V3. The use of DPUs brings an architectural innovation to on-premises data centers that public cloud providers such as AWS (AWS Nitro) have been using for some time. Here, the operating system kernel of the hypervisor is outsourced to the ARM-based CPUs in the DPUs (originating from Nvidia's Mellanox takeover), which on the one hand reduces the load on the Intel/AMD-based CPUs of the host and on the other hand should ensure greater security. In addition, firewall functions (VMware NSX, VMware vDefend) can be processed directly in the DPU. However, initial experience in operating vSphere/VCF with DPUs shows that installation and operational issues such as carrying out firmware updates or reinstalling servers or moving to other clusters become significantly more complicated: Once DPUs are in play, multiple ESXi instances run simultaneously in a DPU-accelerated server. Due to the novelty of the feature and the somewhat increased complexity, administrative teething troubles had to be dealt with until now.

Together with Vmware, Lenovo has come up with a solution to this common problem and introduced a world first for its ThinkAgile VX series servers: dual-DPU operation. This means that two BlueField 2 DPUs can be used in the servers. Customers with demanding application workloads should benefit from either doubled network throughput or failover reliability if one of the two DPUs fails. For this purpose, dual DPU systems can be configured as active/standby, with the standby DPU taking over in the event of a failure.

The manufacturer provides further technical details on dual DPU operation in the announcement. Lenovo pre-configures the dual DPU feature ex works: It is available on the Lenovo ThinkAgile VX650 V3-DPU and requires vSphere 8.0 U3 or newer.

(olb)