Google wants enterprises to use its cloud infrastructure to run AI workloads in or outside its data centers. For the latter, the company made generally available on Wednesday the option of running Google Cloud’s AI platform in a box.
The Google Distributed Cloud (GDC) Edge appliance is mostly for highly regulated organizations that must keep data in-house. The hardware runs the Google Cloud infrastructure stack, data security services and Vertex AI platform. Vertex AI runs models that have been pretrained for various tasks.
Google doesn’t make the hardware. Instead, the company runs its software on products from partners, including Cisco, Hewlett Packard Enterprise and Dell, said Sachin Gupta, Google Cloud general manager for infrastructure services. Nvidia GPUs power the AI models.
“We introduced Google Distributed Cloud as our customers looked at our AI innovations, our data services, our security services, but could not use them in the public cloud,” Gupta said. “[It’s] the cloud in a box that allows you to bring AI anywhere you need it.”
[It’s] the cloud in a box that allows you to bring AI anywhere you need it. Sachin GuptaGeneral manager for infrastructure services, Google Cloud
For example, a Defense Department agency would use the GDC Edge appliance to translate audio and documents written in foreign languages and make the content searchable. To do that, organizations would use the hardware with Google language translation, speech-to-text and optical character recognition software.
Defense manufacturers that are also government-regulated would use the appliance to process data from factory equipment to monitor manufacturing operations without connecting to Google Cloud, Gupta said. The military could also use the appliance to monitor equipment in submarines or aircraft carriers.
The product, introduced last year in preview, has the security clearance for most U.S. military or Defense Department use cases. It has Defense Department Impact Level 5 accreditation for its security features, which include data encryption, data isolation, firewalls and a secure boot that ensures the system will execute only trusted software.
Oil and gas companies would use a ruggedized version of the hardware for monitoring equipment running in remote locations that do not have internet access.
Developers tap the appliance’s computing, storage and networking components and the software running on top through the same APIs used in Google Cloud. The specialized tools and services used for development do require a specific skill set.
“Google treats the edge as a managed extension of the cloud, so it’s most suitable for users who want to build their IoT and edge applications as cloud extensions and who have the skill to develop applications on specialized tools,” said Tom Nolle, principal analyst at Andover Intel, in a report on edge platforms for TechTarget Editorial. “Prospective edge users should take special care in selecting Google’s strategy because of its highly specialized approach.”
IDC analyst Dave McCarthy disagrees that Google is using a specialized approach to GDC Edge. Instead, it contains a cloud-native architecture of microservices, containers and API-driven automation familiar to cloud developers.
“IDC has been tracking that most developers have moved to a cloud-native design strategy,” McCarthy said. “However, there are use cases and workloads that are sensitive to network latency or have specific security compliance requirements.”
Google is not alone in providing edge platforms for running AI and other applications. Cloud providers Amazon, IBM, Microsoft and Oracle have their versions of specialized platforms for distributed computing applications.
In June, Gulf Edge, a subsidiary of Gulf Energy Development, said it would be a managed GDC service provider for Thailand’s utility, energy and healthcare industries. The company planned to focus on providing configurations that did not connect to Google Cloud or the public internet.
Antone Gonsalves is an editor at large for TechTarget Editorial, reporting on industry trends critical to enterprise tech buyers. He has worked in tech journalism for 25 years and is based in San Francisco.
