The cloud computing paradigm has been a table stake for web scale companies for well over a decade, although it’s just now starting to make inroads in the telco industry.
Cloud computing impacts everything. The hardware is commoditized, servers are virtualized, networking and network functions are software-defined. Everything can be managed via software through API calls. Operational challenges are addressed via automation and increasingly applied via artificial intelligence (AI).
Cloud computing also impacts applications and other workloads. They have to adopt to a cloud-friendly design pattern where resiliency and scalability is integral to the application (think stateless microservices). Done right, this transformation results in a more robust and agile platform, enabling a new generation of interactive applications such as connected cars, augmented and virtual reality (AR/VR), and IoT. But done wrong, you are left with a fragile environment and an eroded user experience.
You see, the legacy enterprise stack is built on a fallacy that application server, operating system, and hardware will simply not fail. We all know this to be untrue. This has resulted in a costly operational model where IT is managed reactively and on an emergency basis, which negatively impacts quality of experience as outages are user-impacting. Simply virtualizing legacy workloads and forklifting them to the cloud will aggravate the situation, resulting in more outages and further erosion of user experience. The operational model, as well as the workloads, will have to evolve in order to take advantage of the telco cloud.
The move to 5G and edge computing is exceptionally well aligned with cloud platforms. Consequently, telcos are already phasing out mission-critical servers, instead opting for multiple whitebox servers that can be terminated and replaced with a relaunched server in minutes, without disrupting end-users. In other words, with the hardware becoming commoditized it should also become dispensable (easily replaceable). But, where does that leave the services provided on top of these servers and the end user experience?
Why Do We Want to Have Dispensable Servers?
For a server to be truly dispensable, we should be able to shut it down at will or let it fail without any noticeable end-user impact. We should be able to replace it at some later point, when it is convenient. The new server should join the rest of the system and start executing automatically.
We want this because it changes IT operation from a reactive, emergency operation to a normal business hours function, where the IT engineers pick tickets from their backlog and methodically resolves the issues. And, it is not just the exceptionally positive cost implications, it is the only way to manage the cloud without going bust.
How Do We Get There?
In the telco world, there are multiple environments such as the IT data center, the transport network, and the RAN. Many telcos are already deploying cloud in the IT data center and/or looking at outsourced cloud (e.g. Amazon Web Services – AWS). With the introduction of 5G, cloud is now considered throughout the network. There are many drivers: network slicing and its associated business prospects is just one.
While there are technical challenges, the people and process challenges are about 80 percent of the obstacles to overcome. A bimodal approach to transformation can feasibly soften the transformation. With a new cloud deployed, the focus can be on workloads that are straight forward to migrate to cloud. In the data center, it is mostly about IT applications. At the transport and RAN side, it is mostly about virtual network functions (VNFs). Workloads that are (a) virtualized, (b) disaggregated into discrete functions, and (c) deployable as microservices via continuous integration and continuous deployment (CI/CD) pipelines, are ideal. Workloads that have more traditional enterprise design or are hard-coded to the physical network (e.g. hard-coded IP addresses or assumptions about make and model of a load balancer) stay on the legacy side of the data center (hence, bimodal IT).
People who are adapting to cloud quickly can work on the cloud platform side, allowing staff to gradually adopt the new technologies and processes at their own pace.
What’s the Catch?
Ultimately, the cloud platform is inherently unstable. This is a function of using commodity (cheap) hardware, running cloud-enabled workloads, and responding to business needs via software instead of truck rolls. The agility that is possible comes with a totally new network dynamic. Large internet companies have solved for this, which is why we call them “web scale.” The telco has to achieve the same economy of scale, but is not readily set up to do so. This is where the rubber meets the road.
Changing the profile from an IT and operations culture, to a culture anchored in software and DevOps, is close to impossible within the available window of opportunity. It is instead better to change the vendor relationships, since we are already asking them to de-silo and disaggregate their appliances. A new type of partnership is required where the web scale, DevOps, and open source solutions can be provided with subject matter expertise in a strategic relationship with a new type of “vendor” that is natively cloud software focused.
A telco cloud platform for 5G that facilitates workloads at the edge requires a web scale operational model and a DevOps culture. This is an existential challenge for the telco industry, impacting both people and processes. Partnerships with companies (vendors) can accelerate the transformation. Solutions that facilitate rapid development and deployment of software with a high degree of automation will force change and help people adapt to a DevOps reality. Solutions will have to be cloud friendly, meaning that they scale horizontally and are resilient enough to withstand the instability of the cloud platform. These partners will provide solutions, expertise, and the equally important DevOps culture to the telco customer.
As cloud native implementations proliferate within telco service provider networks, and larger-scale deployment operations are triggered, greater immutability will begin to take shape – with more telcos adopting a similar approach to the solutions of web scale companies. A cloud-enabled infrastructure opens the network to third party applications at the edge, allowing telcos to compete with other cloud providers like Microsoft Azure and AWS. It also enables better utilization and allocation of devices on the network, as it becomes increasingly disaggregated to support future standards. With the right partners and guidance, telcos can build the necessary foundation to deploy low-touch network operations, which are essential in transitioning to 5G.