Sunlight Supports NVIDIA GPU
February 17, 2021
can now unlock the full potential of hyperconverged infrastructure at the edge,
maximizing the performance of demanding, GPU-accelerated workloads such as
artificial intelligence (AI) running in edge environments.
Modern GPUs require massive amounts of throughput to feed the data pipeline.
However, legacy virtualization and cloud infrastructures cannot deliver the
storage IO necessary, resulting in low GPU utilization. They also have a very
large footprint, wasting valuable system resources and leading to high costs and
poor efficiency, particularly in constrained computing environments at the far
Sunlight’s NexVisor drives millions of input-output operations per second (IOPS)
per virtualized instance with ease, allowing 100 percent utilization of GPUs and
enabling high-density, high-performance cloud at a fraction of the cost and
energy consumption offered by legacy virtualization technologies. This is
critical for supporting workloads that demand top performance with the smallest
possible footprint, including AI at the edge.
services provider Yonder Cloud is currently using Sunlight to deliver
high-efficiency GPU-accelerated services. CEO Kevin Smith said, “We run some of
the most demanding workloads for customers on our infrastructure – from AI, deep
learning, visualization and simulation, to VFX, animation and rendering.
Sunlight enables us to achieve full utilization of our NVIDIA GPUs, which
delivers huge cost benefits and efficiency gains.”
“With high-performance AI becoming central to so many business processes, edge
AI has emerged as a very real and rapidly growing phenomenon. Moving AI to the
edge can create massive efficiencies, but it requires a new breed of edge
infrastructure that can process huge volumes of data in the low-power,
resource-constrained and ruggedized compute environments common to far edge
locations,” said Julian Chesterfield, Sunlight Founder and CEO. “Now with
support for super-fast NVIDIA GPUs, Sunlight is uniquely positioned to support
the most demanding workloads out there, with the highest throughput and lowest