Kubernetes-based containerization has revolutionized computing, enabling edge computing and providing a new virtualization alternative for data center computing. Yet, the complexity and limitations of deploying new cloud-native applications using the Kubernetes command-line interface have meant the need to develop Kubernetes skills. In many cases, users decide to keep their current environments, and many applications that could share data and resources remain siloed, especially if the pipeline includes both containers and VMs.
The Rakuten Cloud-Native Platform offers a powerful solution. Through an intuitive app store and extensive automation, it drastically simplifies the creation of sophisticated pipelines and services. I created a short video demonstration focused on using the app store functionality to create these pipelines.
As a Kubernetes-based implementation, the Rakuten Cloud-Native Platform is our Kubernetes-based implementation designed to offer unmatched elasticity, scalability, simplicity, efficiency, availability, and resiliency. It has extensive automation and policy-driven capabilities that make it easy to deploy and manage from day 2 to the end of life.
The applications used in the demonstration video show how it’s possible to create AI pipelines without the Kubernetes complexity and with an intent-based approach.
In the demonstration video, I showcase three pipelines of tools specifically designed to empower data scientists. These tools free them to focus on modeling and tool utilization rather than the intricacies of cloud resource deployment and scaling. This point-and-click simplicity extends to all use cases, making advanced deployments accessible to everyone.
With the Rakuten Cloud-Native Platform, it’s possible to create a pipeline of multiple pods that are managed as a single app. This allows previously siloed applications to share data and reduce network complexity.
For the video, I built three AI pipelines: ai-basics, ai-pro and ai-super. Each pipeline comes with pre-configured with essential AI apps like Jupyter, TensorFlow, Nginx, DataBricks, and PyTorch, providing data scientists with immediate access to a comprehensive suite of resources.
By selecting one of these pipelines from the app store, cloud administrators can facilitate a quick deployment of just the tools they need without needing k8s or programming skills. These pipelines are cloneable and portable to other edge cloud deployments by using the embedded Rakuten Cloud Native Storage functionality. This allows you to take snapshots and use them to back up or clone the entire state of your tools, including data, secrets, and configurations.
The platform also simplifies scaling the pipeline in or out, adding volumes or modifying assigned resources dynamically.
In the video, I demonstrate how pop-up windows show the default resources needed for the pipeline. Users can see the number of instances/pods in use and how many CPU cores, gigabytes of memory and gigabytes of storage are needed to support that pipeline.
Users can slide the scale-out bar to select the number of instances they want, and the interface provides the estimated new CPU core count, memory, and storage resources needed.
Quality of service is managed similarly. CPU core counts and memory usage can be changed to increase the quality of service. Read/write input/output (I/O) operations per second (IOPS) can be adjusted and app priority can be set by clicking on the number of stars (from one to five) needed for this workload.
Once these configurations are set, the environment is ready and can be managed from an app view screen that shows the status of each tool and shows deployment details like IPs, resources, and more.
The tool can also run Day 2 operations for specific pods, such as restarting, stopping, or launching a console.
This view also includes performance KPIs for convenience, which can also be displayed in the embedded Grafana dashboard.
The benefits of Kubernetes containerization get better when pipelines can be created that allow for the sharing of resources and data in a unified cloud capable of running containers and VM-based workloads. Rakuten Cloud-Native Platform makes that easy for both edge compute and data centers. Through its intuitive app store interface and extensive automation, new pipelines can be created in minutes. Get a glimpse of this seamless process in my video demonstration: https://youtu.be/scnlbtBHlGg
.svg)
