IBM Cloud Satellite Makes Your Data Center a Satellite of IBM Cloud

Say you wanted to use an AI algorithm to analyze video footage from a construction site, to make sure all workers are wearing hard hats, or from an office, to see where people congregate the most and focus more cleaning and sanitation efforts there to prevent COVID-19 from spreading. You'd need a lot of processing horsepower to crunch through all that video data, which you wouldn't necessarily have at a construction site or an office.

IBM Cloud and Lumen Technologies (formerly CenturyLink) have been working on solutions for exactly such use cases, deploying IBM Cloud's new variety of hybrid cloud infrastructure in Lumen's edge data centers so that the cloud provider's data crunching tools can process video locally, without transporting huge amounts of data to a public cloud data center far away.

"In a scenario like that you're processing a lot of data," Jason McGee, an IBM VP and CTO for IBM Cloud Platform, told DCK. "You have these live video streams coming off of lots of cameras, and it's hard to pipe all that back to the public cloud and do the processing there, as that would cause a lot of latency issues. What we were able to do is run that workload in the metro area close to the source of the data, do all of the local processing there, reducing latency and allowing us to respond more quickly, but still delivering that as a service through public cloud, with a public cloud kind of operational model."

The new hybrid cloud infrastructure offering is called IBM Cloud Satellite. IBM first announced it last May, during its online Think event, and made it generally available earlier this month. The service offers the ability to extend IBM Cloud capabilities to a data center of the customer's choice, be it their own on-prem facility, a colocation site, or even another cloud provider's platform, such as AWS or Azure.

"To do the video processing, the entire environment was running containerized on an OpenShift cluster," McGee said. "On top of that OpenShift cluster, we had something called IBM Edge Application Manager, which is a set of tools to help you manage devices like cameras and stuff at the edge and deploy applications into the edge. We had a video analytics solution that did the actual video processing and AI recognition of hats or masks, or whatever you're looking for."

"How would you do that traditionally? You would tell somebody to buy a bunch of boxes, install a bunch of software, and you'd have a team of people who had to go to every site."

IBM Cloud Satellite follows in the footsteps of similar offerings by other cloud providers, such as Microsoft Azure Stack and Azure Arc, AWS Outposts, Google Anthos, and Oracle Cloud@Customer. It had been in beta, with participation by more than 65 partners, including Cisco, Dell, Intel, and several telcos, since the announcement last May.

"Satellite essentially is a capability to extend our public cloud services into any location, so customers can consume public cloud in their data center, at the edge of the network, or in other public clouds, like Amazon, Azure, and Google, all managed from the IBM cloud," McGee said. "We can take Kubernetes clusters like OpenShift, databases, or AI capabilities like Watson, and we can deliver them as-a-service anywhere our client needs."

A Public Cloud Experience, Hosted Anywhere

Customers can use their own servers as on-prem Satellite infrastructure, but IBM also offers a 42U rack filled with hardware ready to get the service up and running. It's working with partners like Dell to build integrated systems for customers with more complex needs.

Once the equipment is registered with IBM Cloud, "we kind of take it over and use it as a pool of capacity that we can then run services on. With that capacity, we can reach and run an OpenShift cluster, a database, or an AI framework as a service on that infrastructure in that location, connected back to IBM Cloud and managed by our operations team as a full as-a-service experience, like you would expect in a public cloud. It just happens to live on that infrastructure."

Although the hardware sits locally, it's controlled through the user's IBM Cloud account, which makes life easier for system administrators. If the on-prem cluster loses connection with IBM Cloud (or if IBM's Cloud goes down), the service is designed to continue operating, which can be especially important in edge locations with tenuous network connections.

"You would still consume and manage all the resources from the IBM Cloud," McGee said. "You'd still log into IBM Cloud, but when you define that location on AWS you would have provisioned some VMs in your AWS account, registered those VMs with Satellite, and then we would use that infrastructure to stand up Watson, or stand up OpenShift, or whatever service or application you want to run. But you would interact with all of that through the IBM Cloud."

Besides heavy compute at the edge, an obvious use cases for IBM Cloud Satellite is creating a relatively easy enterprise on-ramp to multi-cloud. Satellite, in this case, would be a staging ground for container deployments across multiple public clouds.

Satellite can also be useful for financial services in countries that prohibit moving citizen data across borders, which complicates use of public cloud platforms.

"We have a financial company we're working with, and they have an online payments platform they white-label and sell to banks around the world," McGee said. "Of course, there are lots of regulations that say the data has to stay in the country that the bank is located in. They can now do that as SaaS in countries where there isn't a public cloud presence, so if they have to go into Russia, Eastern Europe, or someplace where there's not a good public cloud presence, Satellite gives them a way to land a footprint in that country and still deliver it as a service."

Billing for Satellite is no different from billing for IBM Cloud.

"Essentially, there's a small fee for managing the capacity and location," he said. "It's all cloud, so it's a per-CPU-hour kind of metered billing model. Then you pay for the services you're using just like you would on any cloud."