Bitfusion, the Elastic Artificial Intelligence (AI) software company, announces a reference solution architecture combining Bitfusion’s FlexDirect with VMWare and Mellanox platforms for attaching GPUs to any virtual machine over the network.
With the new reference solution, GPU accelerators can now be part of a common infrastructure resource pool and available for use by any virtual machine in the data center in full or partial configurations, attached over the network. The solution works with any type of GPU server and any networking configuration such as TCP, RoCE or InfiniBand. It leverages Bitfusion’s FlexDirect to remotely attach GPUs over the network as well as create fractional GPUs.
“With Bitfusion, VMware vSphere environments can attach any amount of GPU over the network such that they are accessible within any VM in that network,” said Ziv Kalmanovich, senior product manager, VMware. “VMware and Bitfusion enable enterprises to maximize the utilization of GPU infrastructure by forming elastic virtual GPU clusters out of scattered GPU servers in the organization’s network.”
“With Bitfusion along with Mellanox and VMWare, IT can now offer an ability to mix bare metal and virtual machine environments, such that GPUs in any configuration can be attached to any virtual machine in the organization, enabling easy access of GPUs to everyone in the organization,” said Subbu Rama, co-founder and chief product officer, Bitfusion. “IT can now pool together resources and offer an elastic GPU as a service to their organizations.”
“Combining Bitfusion FlexDirect software with Mellanox’s high-performance networking solutions and VMWare vSphere allows customers to consolidate multiple siloed GPU clusters into a single shared platform,” said Motti Beck, senior director, enterprise market development, Mellanox. “Enterprises are looking for higher-performance efficient solutions that maximize the ROI of their GPU based clusters, which the Bitfusion FlexDirect, Mellanox and VMWare joint solution demonstrates, enabling native performance across the suite of benchmarks for network attached full and fractional GPUs.”