What the hybrid cloud’s extension into the enterprise means for AI
Orange Silicon Valley has been experimenting with various architectures for applications and machine learning (artificial intelligence) within our on-premise Kubernetes clusters. We’ve been trying to bridge the gap between the data and applications we host within our enterprise and the advanced services and productivity offered by various cloud computing vendors. Recent endeavors announced by Google and Cisco at the Google Cloud Next ’18 conference in San Francisco (July 24–26) are intriguing and demonstrate how the cloud can extend into the enterprise as a path toward a hybrid cloud.
The hybrid cloud is an amorphous term, often used in enterprise computing to talk about needs related to maintaining on-premise data centers and enabling applications to utilize cloud computing capabilities. Frequently, the perspective comes from within the enterprise and extends into the cloud to access services, creating a wall between on-premise and cloud services. The result is a digital divide between the capabilities available within the cloud and the agility of the local IT teams to procure and deliver similar services.
The consequence of this architecture — or really lack of architecture — is that hybrid cloud applications reach into the cloud in ad hoc ways to access cloud services. Data is pushed from local repositories to cloud services, possibly stored in cloud-enabled data stores, and accessed by compute or machine learning services hosted within the cloud. The result is an unmanaged mesh of on-premise applications and cloud services.
In contrast, the approach presented by Google and Cisco relies on using Kubernetes, a container orchestration technology originating from Google, as a fabric for extending applications and services from the cloud into the enterprise. By deploying and connecting Kubernetes clusters into the enterprise with services similar to those in the cloud, an operator has the same experience through the same cloud management tools. The on-premise clusters just show up as another deployment region upon which application resources can run.
Moreover, services and the rules governing how they interact can be centrally managed, providing a ubiquitous fabric upon which applications can be deployed. Applications deployed within the enterprise can have well-defined access rules between services running in the cloud and vice versa. This allows the operator to see a comprehensive picture of the hybrid cloud, further protecting vital local resources running on-premise.
One key technology for making this happen is Istio — a technology for routing, monitoring, and load-balancing services across various deployments within Kubernetes. (It just released officially as the 1.0 version.) The Istio technology helps operators define who can talk to whom within the cloud into the enterprise — an essential feature required for enterprise security and management. It also provides a uniform way for discovery and routing services.
By reversing the direction and extending the cloud into the enterprise, technologies like Kubernetes and Istio give the enterprise a comprehensive view of the hybrid cloud while allowing managed access between services. The promise is that enterprises can develop strategies where essential customer data and on-premise resources are cloud-connected without compromising the security and protection of essential customer data.
Conceptually, the on-premise data can age into lower-cost storage systems while allowing anonymized and feature-extracted variants of the latest data to be moved to the cloud for analytics and AI purposes. Cloud services can pull from these extracts as they are expected without specializations. This enables the full use of the competitive AI technology stack that is hard, if not now impossible, to replicate on-premise.
In my team’s work and recent developments at Orange Silicon Valley, we see the fundamental promise of Kubernetes as a fabric upon which applications can be deployed, both securely and at scale. The ability for the cloud to extend inwards is a welcomed development that enables our services to utilize the advanced technologies of the cloud while protecting our customer’s data. As a result, we can focus on business innovation utilizing a full array of cloud computing and AI services.
Disclaimer: The views and opinions expressed in this article belong to the author and do not necessarily reflect the position or views of Orange or Orange Silicon Valley.
