Running containers at scale with Kubernetes and Portworx on HPE Synergy

[Ka_Wai_Leung]

Blog by Ka Wai Leung, Solution Product Management, HPE, and Eric Han, VP Product Management, Portworx.

HPE teams with Portworx to offer a reference configuration along with automation to rapidly deploy cloud scale container applications on HPE Synergy.

We live with on-demand applications like Uber and Netflix, where such apps instantaneously respond to our needs. But beyond Uber and Netflix, every enterprise is moving toward cloud-native and container-centric applications. These applications provide enterprises with more agility and greater ability to quickly deliver innovations to your customer base.

Companies looking to adopt and deploy cloud-native or container-centric solutions at scale should consider the HPE Synergy composable platform with Kubernetes and Portworx PX-Enterprise.

This combination leverages:

• Kubernetes as a scale-out and efficient container orchestrator and scheduler
• Portworx PX-Enterprise as the highly reliable, cloud-native storage platform
• HPE Synergy as the composable infrastructure to offer flexible compute and storage to the applications

HPE together with Portworx have encapsulated the best practice of deploying such an architecture in a reference configuration along with deployment automation scripts and playbooks. The outcome of this solution is a robust turnkey approach to running stateful containers.

Containers at scale—optimized for infrastructure

Most enterprise mission-critical applications have need for persistent storage and database access.  Running these “stateful” containers at scale requires compute and storage that are highly efficient, scalable, and available, along with being container-granular. The reference configuration developed by HPE and Portworx, address these needs by using HPE Synergy SY 480 gen10 hardware to provide bare metal compute infrastructure for Docker containers on a Kubernetes cluster. Two HPE Synergy D3940 Storage Modules then provide a total of 80 drives of for Portworx PX-Enterprise to treat as cloud native storage for the containers.

From an application perspective, the HPE Synergy reference configuration for Kubernetes and Portworx PX-Enterprise provides the following benefits:

• Performance—Container applications and backend databases can either run baremetal or on VMs on HPE Synergy SY 480 hardware.
• App services—Kubernetes provides load-balancing, DNS and built-in services.
• Rapid deployment—Synergy image streamer for seamless expansion and contraction of compute and storage resources optimizes resource needs of the container infrastructure.
• Self-healing—Containers get replaced and rescheduled based on server and app-level checks by Kubernetes.
• Highly available data—Portworx creates container-granular volumes that are replicated and kept highly available, including across maintenance operations.
• Container data lifecycle—Container volumes can be automatically encrypted, snapshotted, and backed-up to object storage by Portworx PX-Enterprise.

Rapid time to value

Deploying and configuring a Kubernetes cluster with persistent storage is not easy.  In simplifying the user experience, HPE will automate the entire setup from hardware provisioning through to a running container platform, using a series of Ansible playbooks. The end result is a repeatable configuration for Kubernetes storage to run stateful container workloads with bare metal performance and complete container-level control.

This reference configuration provides best practice guidance for:

• Provisioning compute and storage hardware resources with HPE OneView
• Deploying Kubernetes as part of OS image prepared by HPE Image Streamer
• Launching Portworx PX-Enterprise as a Kubernetes-native daemon
• Running Cassandra (a NoSQL database) as an example of a stateful containerized application

Each Synergy server becomes a Kubernetes worker node that benefits from container-level management by Kubernetes. At the same time, each server is capable of running stateful containers through the Portworx PX-Enterprise storage platform. Applications then run bare metal with storage converged on the same server, which is depicted below.

As part of the built-in capabilities, this Kubernetes and Portworx PX-Enterprise automates the application lifecycle, which further improves the experience for enterprise IT. Applications scheduled by Kubernetes get placed on the optimal server in a scale-out cluster. Afterwards, a set of built-in Kubernetes services, such as DNS and load-balancing, help the application to run and connect to other applications. Kubernetes then manages the application lifecycle for health, for scaling-out, and through upgrades.

Using Portworx, applications do not need to wait for storage to be manually pre-provisioned. Instead, Portworx dynamically creates the required container-granular storage volume to the right size the first time a container runs on a given server. At the same time, IT remains in full control. For example, IT can set different Kubernetes Storage Classes for Portworx volumes, defining at a policy level the requirements for storage’s high-availability and encryption. Portworx then manages the data lifecycle, including taking periodic snapshots and backing up container data to an on-prem or to a public cloud location.

The end-result? Through this solution, IT gets a fully integrated experience for running stateful containers and enterprises get an effective way to run cloud-native on the HPE Synergy platform.

Learn more: HPE Reference Configuration for Kubernetes, Portworx PX-Enterprise on HPE Synergy

Featured articles: 

• How to implement cloud-native computing with Kubernetes
• Comparing composable infrastructure and hyperconverged systems  
• Want to know the future of technology? Sign up for weekly insights and resources 

 Guest blogger Eric Han heads Product Management at Portworx, a storage startup focused on enabling adoption of stateful microservices and cloud-native workflows. At Portworx, Eric has been responsible for the product direction and technical partnerships. Prior, Eric was the first Product Manager for Kubernetes at Google and co-creator of Google Kubernetes Engine.