Covering Scientific & Technical AI | Tuesday, October 8, 2024

Sunlight’s AI Hypervisor Gains Nvidia GPU Support to Boost Edge Deployments 

Sunlight, the U.K.-based specialist in virtualizing data-intensive applications, announced Nvidia GPU support for its “lightweight” hypervisor designed to boost the performance of edge AI deployments.

GPU support for its NexVisor platform would provide “near bare-metal” performance in support of data pipelines currently constrained by legacy virtualization tools and cloud infrastructure, the company said. Those constraints are magnified for data-intensive edge AI applications that struggle with storage I/O performance and other computing resources.

The combination of the streamlined hypervisor used to create and run virtual machines with support from Nvidia (NASDAQ: NVDA) is promoted as delivering the performance upgrade needed to feed data pipelines. It is also touted as tapping the potential of hyperconverged infrastructure at the edge while boosting the performance of GPU-accelerated AI workloads running on edge networks.

The result would be full utilization of GPU performance via NextVisor, which is billed as providing millions of IOPS per virtualized instance. For edge applications, Cambridge-based Sunlight said that capability would translate into higher throughput and lower latency for edge AI workloads.

The GPU-backed hypervisor is currently used by Sunlight’s customers for machine learning, simulation and visualization workloads along with visual effects and rendering.

The GPU deal is the latest example of how GPU acceleration is expanding throughout big data processing frameworks. For example, Nvidia announced support last year for native acceleration of Apache Spark. Running Spark workloads on GPUs would help simplify data pipelines while accelerating the training of machine learning models, Nvidia said last May in announcing native support for Spark 3.0.

Last summer, Nvidia unveiled its GPUDirect Storage technology aimed at leveraging GPU performance by creating a direct path between local and remote storage, thereby overcoming I/O bottlenecks that are slowing the processing of AI and HPC data sets.

Those approaches now gain a hypervisor from partner Sunlight designed specifically for edge AI deployments that require a smaller footprint, according to the hyperconverged infrastructure stack vendor

Julian Chesterfield, Sunlight’s founder and CEO, was one of the architects of Xen virtualization, a foundational component of Amazon Web Services (NASDAQ: AMZN), among other cloud vendors. The five-year-old startup announced a $6 million Series A funding round this past December led by OpenOcean along with participation from Robert Bosch Venture Capital.

This article first appeared on sister website Datanami. 

About the author: George Leopold

George Leopold has written about science and technology for more than 30 years, focusing on electronics and aerospace technology. He previously served as executive editor of Electronic Engineering Times. Leopold is the author of "Calculated Risk: The Supersonic Life and Times of Gus Grissom" (Purdue University Press, 2016).

AIwire