Today, Radium, a startup that aims to harness artificial intelligence and machine learning to extract more computing power from cloud hardware, announced it is exiting stealth mode and deploying its solutions in Cyxtera-operated cloud data centers in Toronto, the New York and New Jersey metropolitan area, and Silicon Valley.
The main product, called Launchpad, allows users to start and shut down projects on bare metal machines, eliminating the need for extra layers of hypervisor or virtualization software. Radium provides benchmark tests for machine learning jobs and found speed improvements ranging from 30% to 140%.
“Our initial testing shows that bare metal servers provide an excellent cloud computing platform for the high-performance deep learning and inference workloads required for these types of applications,” said Srinivasa Narasimhan, a professor at Carnegie Mellon University’s School of Computer Science, who has worked with the company to test the product.
Improving hypervisor performance
Many cloud products rely heavily on a virtualization software layer, or “hypervisor,” that allows a single physical machine to simulate many smaller machines that appear independent to the user. However, implementing a hypervisor is costly. Simulations often add small delays to tasks such as retrieving data from disk or sending data packets over the Internet.
[ Related: The emergence of datacenter-as-a-service ]
“Broadly speaking, we felt that the diluted nature of middleware and hypervisor stacks within major cloud provider architectures could be improved,” said Radium CEO and co-founder Adam Hendin.
Radium’s bare metal approach is optimized to target the AI market in research institutions and enterprise stacks. The smart deployment software built into Launchpad is designed to help you optimize the performance of the AI algorithms running inside.
This product is priced with data-intensive workloads common in the field in mind. There are no additional charges for ingress and egress. This pricing set is often overlooked, but it can sometimes be a big surprise to users of other clouds.
Hear from the ML community
“We designed and purpose-built a cloud platform that exceeds the needs of the most demanding AI workloads,” Hendin explained. “A public cloud with the benefits of a private cloud. We have also listened to the ML community who are frustrated with non-portable software and high data egress charges within the large cloud providers that lock customers in.”
Other companies are experimenting with using AI algorithms to allocate resources in the public cloud. For example, Zesty offers a product that monitors instances in public clouds such as AWS and makes decisions to save money by shrinking or even shutting down over-provisioned machines.
[ Related: Cloud datacenters anticipated to become largely robot-dependent by 2025 ]
Cyxtera already operates 61 data centers in 21 markets around the world. The company primarily specializes in colocation and bare metal servers for the enterprise market, but has also expanded to offer services such as AI/ML training using Nvidia machines.
Deploy Radium in your data center
“Cyxtera’s global platform and scalable interconnect solutions provide Radium and its customers with a reliable foundation to grow their business as their workloads expand,” said Holland Barry, Cyxtera’s senior vice president and field CTO.
Cyxtera has deployed the current version of Radium Launchpad in three centers so far, with plans to expand soon.
“Radium is breaking the mold of cloud automation with its Launchpad ecosystem, enabling a simple process for provisioning, configuring, and deploying,” said Barry.
