Comments
Virtualization

Container Workloads: Evolution of SNAPS for Cloud-Native Development

Feb 8, 2018

Application developers drive cloud-platform innovation by continuously pushing the envelope when it comes to defining requirements for the underlying platform. In the emerging application programming interface (API) and algorithm economy, developers are leveraging a variety of tools and already-built services to rapidly create new applications. Edge computing and Internet-of-Things (IoT) use cases require platforms that can be used to offload computing from low-power devices to powerful servers. Application developers deliver their software in iterations where user feedback is critical for product evolution. This requires building platforms that allow developers to develop new features rapidly and deploy them in production. In other words, to adopt DevOps.

In the telecommunications world, network function virtualization (NFV) is driving the evolution of telco clouds. However, the focus is shifting towards containers as a lightweight virtualization option that caters to the application developer’s requirements of agility and flexibility. Containerization and cluster-management technologies such as Docker and Kubernetes are becoming popular alternatives for tenant, network and application isolation at higher performance and lower overhead levels.

Container is an operating system level virtualization that allows execution of lightweight independent instances of isolated resources on a single Linux instance. Container implementation like Docker avoids the overhead and maintenance of virtual machines and helps in enabling portability and flexibility of applications across public and private cloud infrastructure.

Microservice architectures are enabling developers to easily adopt the API and algorithm economy. It has become imperative that we start to look at containers as an enabler for carrier-grade platforms to power new cloud-native applications.

Edge computing and IoT require containers

Edge Computing and IoT are introducing new use cases that demand low-latency networks. Robotics, autonomous cars, drones, connected living, industrial automation and eHealth are just some of the areas where either low latency is required, or a large amount of data needs to be ingested and processed. Due to the physical distance between the device and public clouds, the viability of these applications depends on the availability of a cloud platform at the edge of the network. This can help operators and MSOs leverage their low-latency access networks—their beachfront property—to enable such applications and create new revenue streams. The edge platforms require cloud-native software stacks to help “cloud-first” developers travel deep inside the operators’ networks and make the transition frictionless.

On the other hand, the devices also require client software, which can communicate with the “edge.” The diversity of such devices such as drones, sensors or cars makes it difficult to install and configure software. Containers can make life easier since they require a version of Linux operating system and container runtime to launch, manage, configure and upgrade software to any device.

The role of intelligence and serverless architectures in the carrier-grade platform

Let’s consider the example of a potential new service for real-time object recognition. By integrating artificial intelligence (AI) and machine learning (ML) algorithms, operators can enhance the edge platform so developers can create applications for pedestrian or obstacle detection in autonomous driving, intrusion detection in video surveillance and image and video search. The operator’s platform that hosts such applications needs to be “intelligent” to provide autonomous services. It requires the ability to host ML tools and support event-driven architectures where computing can be offloaded to the edge on-demand. Modern serverless architectures could be a potential solution for such requirements, but containers and cloud-native architectures are a near-perfect fit.

Are containers ready for carrier-grade workloads?

Containers as a technology have existed for over a decade. Linux containers and FreeBSD Jails are two early examples. However, it was not easy to network or manage the lifecycle of containers. Docker made this possible by simplifying container management and operations, which led to the ability to scale and port applications through containers. Today, the Open Container Initiative of the Linux Foundation is defining the standards for container runtime and image formats. APIs provided by container runtimes and additional tools help abstract low-level resource management of the environment for application developers. Container runtimes can download, verify and run containerized application images.

The production applications are typically composed of several containers that can independently scale. To manage such deployments, new software ecosystems have emerged that primarily orchestrate, manage and monitor applications across multiple hosts. Kubernetes and Docker Swarm are examples of such solutions, commonly called container orchestration engines (COE).

Some of the key challenges for carrier-grade deployments of container-based platforms are:

  • Complex networking with several alternatives for overlay and underlay networks within a cluster of containers
  • Lack of well-defined resource-management procedures like isolating containers with huge pages, CPU pinning, GPU sharing, inter-POD, node-affinity, etc.
  • Complex deployment techniques are required to deploy multi-homed PODs
  • Large ecosystems for securing container platforms as it is not easy to deploy and manage large container security solutions

SNAPS and Containers

SNAPS, which is short for SDN/NFV Application Development Platform and Stack, is an open-source platform developed by CableLabs. The platform enables rapid deployment of virtualized platforms for developers. SNAPS accelerates adoption of virtual network functions by bootstrapping and configuring a cloud platform for developers so they can focus on their applications. Aricent is involved in the SNAPS-OpenStack and SNAPS-Boot projects and contributed to the platform development with CableLabs.

An obvious next phase is to enable containerized platforms. A key first step was already achieved in the SNAPS-OpenStack project where Docker containers are used for executing many OpenStack components. The next obvious step is to create a roadmap for enabling containers for application developers. A cursory look at the cloud-native landscape reveals that this ecosystem is huge. There are several options available for DevOps, tooling, analytics, management, orchestration, security, serverless, etc. This can create confusion for developers regarding what to use and how to configure these components. They will have to “learn” the ecosystem, which will delay their own application development. The future roadmap for SNAPS is to enable developers by bootstrapping a secure and self-service container platform with the following features:

  • Container orchestration and resource management
  • In-built tooling for monitoring and diagnostics
  • A reference microservices architecture for application development
  • Easy management and deployment of container networking
  • Pre-configured and provisioned security components
  • DevOps-enabled for rapid development and continuous deployment

These are exciting times for developers. The availability of platforms and technologies will drive innovation throughout the developer community. The SNAPS community is focused on ensuring that the best-in-class developer platforms are created in the spirit of open innovation. The SNAPS platform roadmap adopting cloud-native ecosystem is going to provide developers an easy-to-use platform. We are looking forward to a larger participation for the developer and operator community. As a community, we must solve the key challenges and create a resilient platform for containerized application platform for network applications.

Have Questions? We’d love to hear from you: 

--

The author, Shamik Mishra, is the Assistant VP of Technology at Aricent. SNAPS, CableLabs’ SDN/NFV Application Development Platform and Stack Project, was developed leveraging the broader industry’s open source projects with the help of the Software Engineering team at Aricent. CableLabs selected Aricent for this specific project because of their world-class expertise in software-defined networks and network virtualization. In a little less than a year, CableLabs and Aricent worked closely to extend CableLabs’ initial code base to the full SNAPS platform. The SNAPS platform has now been released to open source to enable the wider industry to collaboratively build on our work and to use it to test new network approaches based on SDN and NFV.

Comments