With the transformation from hardware-centric networks to networked cloud platforms, the relationship between telecommunications companies (telcos) and their server(s) is inherently changing. Cloud computing introduced a software design paradigm that required a change in the ways enterprise and telco approach operations, and with it came a new way of thinking. This innovative shift can be conceptualized using the pets vs. cattle analogy. Initially coined by Bill Baker, a veteran software developer, the analogy was later applied to cloud computing by Randy Bias, VP, technology and strategy at Juniper Networks. With the dawn of modernized telco networks and network clouds, the pets vs. cattle analogy is more relevant than ever.
A traditional, monolithic architecture is like your pet dog Spot. Spot is unique and irreplaceable – you named him, feed him and care for him when he’s sick. You might even say he’s mission-critical to your life. With the mission-critical servers on racks, if one goes down, the end-user is impacted. These types of “pet” servers are being phased out in favor of dispensable, horizontally scaled servers.
In modern, cloud-based and microservices software systems, infrastructure is the cattle, and you are the IT rancher. Your relationship with your herd of cattle differs greatly from your relationship with Spot – the cattle requires little human intervention, and when one gets sick, it’s easily replaced with another. IT professionals are opting for homogenous, commodity “cattle” servers, running stateless microservices that can be terminated and replaced with a relaunched server in seconds, without impacting end-users.
Transition to modern, microservices-based infrastructure in telecoms—three tenets of success
While network transformation generally refers to microservices workload architectures, many vendors are only offering virtualized appliances – These are pets masked as cattle. If the infrastructure has been virtualized, but you can still call it an appliance, it’s a pet, albeit capable of being housed on commodity hardware. Worse, this pet is now more fragile than before and your objective to transform to a cloud platform has failed. The transition to virtualized network functions is truly about “function”. If the word “appliance” is still part of its description, you should be alarmed.
This is important for operators as the move to the 5G network impacts function. Having a cloud infrastructure capable of launching virtual machines and/or containers, is a prerequisite, but not enough on its own. There are three technical tenets of a successful cloud transformation to a true cloud platform: (1) true virtualization for cloud; (2) software-managed cloud and (3) life-cycle-management (LCM) of cloud. True virtualization for cloud means any workload running in the cloud must be designed for cloud. Merely virtualizing an application and delivering it as a “virtualized appliance,” will only let it run on commodity hardware at less than 99.999% availability (a.k.a, “five nines”). The application must consist of stateless microservices, and other cloud-friendly constructs, for it to scale in a robust way.
A software-managed cloud will allow you to move to a “software-defined” business where you can instantly respond to business needs. This means the BSS and OSS interacts with the cloud via its API. Additionally, the cloud platform should be controlled and managed automatically via software only – no more ticket systems with IT staff manually implementing requests or human monitoring and controlling the networks 24/7 in real-time. Life-Cycle Management (LCM) on a cloud platform is also called Continuous Integration/Continuous Deployment (CI/CD). Without it, you cannot respond to business needs fast enough. Instead of truck rolls or manually deployed hardware, we now update, upgrade and migrate all things running in the network via software. While this has the promise of accelerating IT processes, it requires the CI/CD pipeline to conduct LCM. Other factors that come into play include organizational structures, culture, strategy and business objectives. However, if we cannot overcome the technical aspects, nothing else matters.
Bringing cloud to the edge—becoming a rancher of many ranches
As operators continue to drive toward 5G, cloud platforms will be highly distributed and made up of smaller, inter-connected edge clouds. To date, cloud-native applications have been running within the data center. They are made up of largely stateless and idempotent microservices, inter-connected via a messaging mechanism, such as a message bus (e.g., RabbitMQ or ZeroMQ), allowing the developer to build resiliency and scalability into the application. In the telco environment, each edge cloud hosts applications and other workloads (i.e. Virtual Network Functions). The universe of the horizontally scaling application now takes a new and larger meaning. In addition to resiliency and scaling for load, the application needs to scale for latency, at the right location. Doing this right, will enable Virtual Reality (VR), Augmented Reality (AR), drones, self-driving cars, interactive gaming and other real-time applications.
Preparing for 5G isn’t the only benefit to this approach, however, microservices-based infrastructures allow for faster innovation, cost savings, increased service revenue, maintained performance/resilience and end-to-end orchestration and automation (1). Cloud enabled infrastructure opens the network to third party applications at the edge, enabling telcos to compete with other cloud providers like Azure and AWS. It also allows for better utilization and allocation of devices on the network.
With traditional infrastructure, you don’t have elasticity, and can’t manage capacity in a dynamic way, because you’re still deploying a pet. A tremendous time and cost saving benefit of scaling horizontally is the ability to increase capacity of serves at a moment’s notice. Swapping out one cattle for another is fast and, rather than spending money on costly specialized hardware, administrators can add as many inexpensive commodity servers to a cluster as desired (2). In addition, the system is managed over software and can be remotely and uniformly updated.
As a telco provider in today’s environment, it is imperative that any mission critical network is truly cloud native. Although Spot can be masked or marketed as a virtualized appliance, he’s still a pet, and largely incompatible with a true cloud platform. Unless you can conclude that the virtualized infrastructure and network functions is microservices-based and looks like cattle, you have achieved nothing. For edge cloud, horizontal scaling for capacity and resiliency is only part of the equation and knowing where to scale is equally important. The telecoms industry is undergoing a transformational change, and network providers who are proactively embracing these changes will likely have the most efficient and effective telco ranch.
