If you live someplace warm, you’re probably used to cranking the air conditioning on hot summer days. Sure, AC uses lots of electricity, but life can be unbearable without it. But what if the technical constraints of air conditioners required them to stay fully powered all year long—burning the same electricity month after month, even when they’re not turned on? That would be a huge waste of money, not to mention the effect on the environment. Unfortunately, most telecom networks still operate on a similar model.
Telecom infrastructure utilization can fluctuate wildly with demand. To assure reliable service, Communication Service Providers (CSPs) traditionally have overprovisioned their networks for peak usage—even though most of the time, just a fraction of that capacity is needed. This means that for most of their service lives, telecom networks continue to consume power—and generate CO2 emissions—for reserve resources that aren’t even used.
Now, cloud-native 5G networks promise a new era in CSP sustainability. Containerized 5G infrastructure can scale in and out with demand, activating—and powering—only the network resources needed at a given time. This elastic scalability could also deliver huge cost savings. Just as important for CSPs committing to Environmental, Social, and Governance (ESG) targets, it can make a major dent in the network’s carbon footprint. At least, in theory. In practice, auto-scaling cloud-native 5G networks is more difficult than it sounds.
The problem: telecom workloads come with unique challenges that most cloud users never have to worry about, and that standard cloud orchestration mechanisms don’t address. If you’re with a CSP organization seeking to tap into cloud scalability, make sure you’re thinking through these issues carefully. Otherwise, you could end up with a network that doesn’t deliver the ESG benefits you expect—and carries much higher costs than planned.
Unleashing telco sustainability
CSPs are always looking to reduce operating costs, but improving sustainability has become equally important. For some, attaining ESG certification (and eligibility for ESG investment) is important for long-term financial health. Others simply want customers to know that they share their commitment to a greener future. Regardless, CSPs in every market are launching ambitious initiatives to reduce carbon footprint and move towards “net-zero” emissions.
Meanwhile, CSPs are also adopting cloud-native architectures in preparation for 5G Standalone (5G SA) networks. These trends couldn’t converge at a better time. By replacing monolithic virtual machine and physical network appliances with independently scalable microservices, cloud-native architectures can potentially deliver huge improvements in power efficiency (Figure 1). Instead of overprovisioning, CSPs can dynamically scale out containerized 5G network functions (CNFs) when usage spikes, and then scale back in when it declines—consuming only the energy needed to satisfy demand.
Figure 1. Modern Cloud Evolution
Cloud auto-scaling isn’t new; hyperscalers have used it for years in the world’s largest data centers. Even previous CSP virtualization efforts, like Network Functions Virtualization (NFV), could theoretically support it, although in practice it was rarely deployed due to inconsistent vendor support. Now though, 5G promises to bring true elastic scalability to telecom. Partly, that’s because cloudification is no longer optional; CSPs must adopt microservices-based architectures to support 5G SA. More practically though, cloud-native architectures are mature, validated, and controllable via standard tools like Kubernetes.
The ability to use proven, industry-standard orchestration mechanisms should help CSPs implement cloud technologies more successfully. But there’s one catch: standard auto-scaling mechanisms were designed for enterprise workloads, not telecom. And as some CSPs have discovered, you can’t just apply the same approaches and expect the same result.
What’s different about telecom?
Most cloud-native environments—whether enterprise data centers or massive public clouds—are optimized for web transactions, which tend to be shorter-lived and relatively fault-tolerant. (If a web page loses connectivity, you can just reload the browser and carry on.) That’s often not the case for telecom.
CSP workloads often have a real-time component, making failures more noticeable and disruptive to the user experience. Additionally, 5G nodes can support hundreds of thousands of highly concentrated users. If a network element fails, it doesn’t just annoy a few subscribers; entire regions can be affected. The biggest barrier to 5G auto-scaling, however, is just finding effective mechanisms to scale in CNFs.
This can sound counter-intuitive, as scaling out—activating new resources on demand—seems like it would be more difficult. But consider: before powering down unneeded 5G resources, all subscribers must be moved off that CNF. Unlike with web transactions, this process can take a long time, and it’s difficult to move real-time sessions without breaking them. Suddenly, those proven mechanisms that CSPs are counting on to scale in their networks won’t work.
The power of fine-tuning 5G networks
Cloud-native 5G networks can dynamically scale in, even in complex real-time scenarios. But it doesn’t happen “out of the box”—at least not today. For now, auto-scaling requires extensive fine-tuning of both Kubernetes and CNFs in an iterative, trial-and-error process.
The good news is that 5G CNFs and cloud infrastructures are highly configurable, allowing you to adjust thresholds, timers, and every aspect of dynamic scaling. But in order to tweak configurations, measure, and fine-tune, you have to be able to visualize the cloud layer, 5G services layer, and user experience all at once. This requires full visibility and expansive testing capabilities. You need to be able to:
- Validate that cloud infrastructure performance can support diverse, distributed 5G workloads
- Proactively test the dynamic scalability and resiliency of 5G CNFs and cloud infrastructure, including with realistic traffic that varies over time
- Quickly troubleshoot causes of performance issues in highly dynamic, multivendor cloud environments
In some early deployments, CSPs without these capabilities found their cloud-native 5G networks unable to scale-in. The anticipated power and CO2 reductions didn’t materialize. And if they were running 5G workloads in public cloud, the consumption-based price tag was much higher than expected.
Eventually, CSPs and their vendors will develop industry-standard auto-scaling mechanisms for telecom. With some CSPs estimating that they currently consume 4-5x the electricity they’d need in an optimized environment, the opportunity is just too great to ignore. For now though, make sure you’re building the comprehensive testing capabilities needed to deploy 5G auto-scaling successfully, or working with a partner who can provide them for you. Your customers—and the planet—will appreciate it.
