The concept of computing container technology has been around a few decades, depending on what you consider a "container." In the past decade, enterprise IT and web-scale cloud providers have begun to rely upon containers to run many applications, starting up billions of them every week.
At the same time, we have also seen many industries being disrupted or impacted by personalized digital services. The world is abuzz with how 5G and edge computing will change our lives, maybe even more than the introduction of the smartphone. Before we leap forward into 5G, let’s catch up on what communications service providers (CSPs) and our industry have been able to accomplish with current infrastructure.
Our CSP customers like Turkcell have been actively building private clouds for the past few years to accelerate and automate the deployment of their mobile (4G/LTE) and business (SD-WAN) network services using network function virtualization (NFV) to package those applications in virtual machines (VMs) on off-the shelf server hardware.
In parallel, some have collaborated with their technology and business partners to embrace the use of containers, microservices architectures, and cloud-native applications. Our CSP customers have been deploying these technologies on Red Hat OpenShift Container Platform in production across a diverse set of IT, network operations, video, and OSS/BSS application environments to modernize their businesses and accelerate time to revenue.
As one might expect, more of our CSP customers would like to gain the business benefits of containers and cloud-native architectures provided by Red Hat OpenShift by expanding further across their IT, operations, and data applications and into their rapidly changing networking environments. From a network applications perspective, containers, microservices, and cloud native approaches offer compelling potential benefits compared with existing VM environments:
Lower overhead: Containers consume fewer hardware resources for less time than virtual machines because they have a smaller memory footprint as a result of virtualizing the operating system (OS) rather than the hardware. Plus, service and application designers can use containers to initiate and process individual transactions or sessions and release the container resources as soon as those complete.
Faster startup times: The time taken to start a new VM is measured in seconds, primarily driven by moving the image onto the host system. Container images are small applications modules and can start up in fewer than 50 milliseconds in many cases.
Ease of use: Containers provide a high degree of portability across operating environments, making it easy to move workloads between private and public clouds. Cloud-native applications – modular collections of microservices packaged in containers – can provide elasticity and scale for each function, making them easier to deploy.
Reduced maintenance: Compared to virtual machines running a variety of operating systems, maintaining applications in containers can be much easier. If you're using something like Red Hat OpenShift Container Platform with RHEL base images for your applications, you can apply updates and roll them out in an automated fashion with little effort. Compare this to patching each virtual machine's operating system, and you'll find containers (when used correctly with the right tools) can require much less effort to maintain.
NFV has matured significantly in recent years, moving key functions from physical to virtual with many production deployments of mobile packet core, firewalls, policy, deep packet inspection, and video optimization from their technology partners. CSPs have accelerated their service deployment times with automation and reduced their operational costs while supporting hundreds of millions of subscribers around the world using horizontal cloud platforms based on Red Hat OpenStack Platform.
So how does container and cloud-native technology fit into the next generation 5G mobile infrastructure announcements and plans of communications service providers? CSPs see more than just higher speed radio access networks (RANs) with 5G. They also see the opportunity to deploy innovative new applications and services for their enterprise customers on top of that new highway to the edge. CSPs are looking at containers and cloud-native technologies to play a significant role in optimizing and mitigating the complexities of their distributed 5G/edge mobile infrastructures.
Sounds exciting, doesn’t it? What’s the catch, and when will this occur? Well, there are a few key challenges facing the industry that we are all working together to overcome.
CSPs have deployed large volumes of network applications in both physical and, more recently, virtual instances across a wide range of functions from their technology partners. The complexity of decomposing these applications into modular microservices to place into containers requires some heavy thinking and effort to maximize the gain from modernization. As a result, the mileage will vary for both development cost and time to completion.
The industry has spent considerable effort in maturing NFV deployments on OpenStack to support capabilities many of us took for granted with physical networking elements – capabilities from multi-tenancy, load balancing, security, forwarding performance, to integration with software defined networks and routing, just to name a few. Expanding to the RAN and to the scale of 5G requires support for real-time applications, synchronization, IPV6, and distributed computing in the mix.
Much like taking advantage of virtualization through NFV, service providers are now looking to bring cloud-native, containerized application benefits to network functions. Kubernetes has become the default open source project that controls the deployment, scaling, automation, and management of containerized applications. While its simple networking model is well-suited for web applications and microservices, it does not, however, currently suffice for high-performance virtual network function (VNF) applications that must deliver higher service level objectives (SLOs).
A large number of open source projects and working groups are in full swing to complement Kubernetes (e.g. container network interface, Multus, resource management, network) to meet the robust, secure, and operational management needs of deploying cloud-native networking applications for service providers. To give you a sense of the pace of innovation, 95% of the 3 million lines of code for Kubernetes itself have changed in just the past few years.
Linux containers are portable between cloud environments and you can run your application on premise or on a public cloud. However a key misconception about containerizing applications is that you can mix and match them and run them on any Kubernetes-compliant container management platform.
While OCI defines a standardized interface between the OCI-compliant container images and the OCI-compliant container engines (like CRI-O), it does not let us see how an application interacts with the underlying kernel, such as visibility to any of the hundreds of system calls just to start up and run it.
So, what is the current state of the art in the industry today? Our CSP customers like Telenor and more than 50+ others continue to expand their deployments of NFV on Red Hat OpenStack Platform.
At the same time, Red Hat OpenShift has become the open platform of choice for many of our customers across all industries from financial (BBVA), automotive (BMW), healthcare (HCA), and to service providers like Telstra, that continue to expand their deployments of cloud-native applications both on-premise, on multiple public clouds, and as-a-service.
Containerizing a set of services like those in the Open Network Automation Platform (ONAP) and running them on a supported container management platform like Red Hat OpenShift has provided valuable insights into the use of container and cloud-native technologies by our service provider customers.
We also have customers like X by Orange that are deploying cloud-native business security service and music streaming applications in combination with network functions from their technology partners that are still packaged in VMs, fully automated and deployed with Red Hat OpenShift and delivered on Amazon Web Services public cloud.
Our global ecosystem of technology partners are working very closely with Red Hat and our CSP customers to accelerate the modernization of their next generation 5G core and RAN applications, along with many others, to deploy them in a cloud-native manner on Red Hat OpenShift.
Will virtual machines disappear? No. Not anytime soon. Just like we still see continued deployments of physical network elements (especially in optical, radio, access, and routing), we will continue to see virtual network functions (VNFs) deployed by our CSP customers on Red Hat OpenStack Platform for some time. Virtual machines will evolve to become part of a cloud native fabric combined with containers and microservices rather than in a separate VM based environment.
We do not see our CSP customers quickly changing out technologies deployed in production. It is operationally cost prohibitive and they wish to maximize the return on those investments. Many CSPs cap older technologies and deploy newer technologies for the growth and expansion of their infrastructure and services to mitigate their risk and costs.
We can expect to see the same approach for the evolution from virtual machines to containers and cloud native deployments as we evolve from 4G/LTE to next generation 5G and edge computing services. With Red Hat OpenShift, our CSP customers have the freedom to develop and deploy their applications, along with those from our technology partners in multiple ways and on the cloud infrastructure that meets their needs.
Our CSP customers may deploy some applications on bare-metal, others as VNFs in VMs packaged in containers, and still others as modular cloud-native network function (CNF) microservices applications. Some of those services may be stateful, some may be stateless. How you manage all that, break it all down into individual components, automate, and efficiently operationalize different services is really what next generation 5G/Edge cloud native infrastructure is all about.
If you would like to hear more on this topic, make sure to watch a replay of my recent webinar: Accelerating The Journey: Cloud-Native Architectures For Digital Service Providers.
About the author
Ian is Chief Technologist, Global Service Provider Business at Red Hat, the world’s leading provider of open source technologies. Ian brings more than 30 years of engineering, business, and telecommunications industry leadership to Red Hat. Acting as a catalyst and trusted advisor, Ian brings together a wealth of industry and open source community insight to help our customers flourish in the digital economy.