Three principles for bringing hyperscale cloud models to service provider networks (Reader Forum)
“Cloud is eating the world.” At least, that’s the big takeaway from the tech industry press these days. And that’s a good thing. For consumers, cloud means being able to access digital services whenever and wherever you choose, without having to worry about how it all happens behind the scenes. But what if you’re responsible for making everything happen behind the scenes? If you’re part of a service provider organization, what does cloudification actually look like?
Broadly, you can expect data center (DC) architectures to become the basic template for an ever-larger portion of your portfolio. The good news is that cloud DC approaches are among the most mature and proven in the industry, pioneered by hyperscale leaders like Google, Microsoft and Amazon. Now, these powerful cloud concepts can play a much larger role in your business.
What does the growing reliance on cloud DCs mean for service provider operations teams? And what should you be thinking about as you navigate this transition? Let’s take a closer look.
Understanding service provider data centers
Data centers have played a role in both internal and customer-facing service provider operations for years. The difference is that now, more of the subscriber network itself — from core to edge to radio access network (RAN) — is getting virtualized. As more of your network becomes software running on commercial off-the-shelf (COTS) servers, cloud DC concepts become applicable to more and more of what you do.
But DCs are not monolithic. Today, service providers use DCs in three main ways:
- IT data center: Functionally, service providers are large enterprises, and like any business, they need to support a variety of enterprise applications such as Salesforce, Microsoft 365 and many others. However, service provider DCs also support applications that directly affect subscribers, such as operational support systems and business support systems (OSS/BSS). So, even as more applications move to public cloud, IT DC infrastructure will continue to play an integral role in the business.
- Managed services data center: Service providers use theseDCs to provide managed services to their (usually enterprise) customers. That can include things like managed web security, software-defined wide-area networking (SD-WAN), secure access service edge (SASE) products and others.
- Telco cloud/edge cloud data center: This is the area seeing the biggest change, as operators adopt Telco Cloud architectures to support foundational telco workloads. In core networks, that includes virtualizing and cloudifying functions like evolved packet core (EPC) and IP multimedia subsystem (IMS). At the edge, operators are cloudifying RAN sites to enable Open RAN concepts, as well as new edge compute-based enterprise services such as content delivery networks, connected factories and many others.
We’re not yet at the point where operators are converging these different DCs onto a single fabric and we may never get there. These DCs can run vastly different types of workloads with vastly different requirements necessitating a variety of DC topologies and presenting CSPs with imposing operational challenges. While designs can and should vary, common DC fabric management and automation systems, along with software-defined networking tools, can bring operational consistency that delivers the business agility that CSPs are striving for. This means you can now take a page from hyperscalers, and replace yesterday’s slow, inefficient operational processes with modern cloud approaches.
Key principles for scaling your data center footprint
Cloud approaches promise big benefits. When you can run your network as a pool of flexible cloud resources, you can roll out new services much more quickly. You can deliver better performance and reliability to customers, and you can operate your network more efficiently, at a lower cost. At least, modern cloud approaches can deliver these benefits. But just because you’re running more of your network like a DC, that doesn’t mean you’ll automatically realize them.
As you plan your cloud investments for the coming years, keep the following principles in mind:
- Build with an experience-first mindset: As you deliver a more demanding mix of services, cloud lets you move capacity wherever it’s needed more easily. But capacity only takes you so far. Especially for telco/edge cloud operations, it’s less about how many bits you can push per second and more about the experience you deliver. This means not only the experience for your end users, but also the experience encountered by your DC operations teams. Day 0/1/2 lifecycle management can’t be an afterthought. But employing infrastructure-as-code tools to speed up DC planning, design and configuration is only the beginning. The real savings come with simplifying ongoing operations. Ensuring operations teams can consistently and reliably manage infrastructure and deliver great experiences to customers, whether internal or external, needs to be the primary focus of every design decision you make.
- Intent-based automation is now essential: Service provider networks are getting more complex, and new 5G and edge services make them exponentially more so. The only way to tame that complexity is with automation. But too many automation approaches focus on accelerating antiquated manual processes rather than replacing them. Yes, you can use scripting to speed up CLI configurations, but human beings still have to figure out the step-by-step techniques for those scripts. To run your business more like a hyperscaler, you need to abandon that model altogether. Instead, you need intent-based automation —automation tools that use AI engines and machine learning to take in vast amounts of real-time information about your network. The operational mindset must shift from job-to-be-done to decision-to-be-made. Telemetry based on graph databases, which excel at correlating massive amounts of information, help with this transition. Furthermore, state-based approaches, rather than event-based systems, also help sharpen decision-making.
- Prioritize operational consistency: One of the biggest benefits of 5G is its expansion of open standards and APIs, allowing service providers to introduce more vendors to more parts of the network. This is great news—more vendors means more competition, more innovation and better economics. But if you’re not careful, it can also mean massive operational complexity when different parts of the network require different vendor-specific tools and processes. Make sure that no matter how many vendors you bring to your environment, you can use the same consistent operations. Your DC automation tools should be able to manage infrastructure from any vendor and even white box with ease. Operational consistency has been a cornerstone of hyperscaler strategy and a key to their success. Consistency leads to reliability. Reliability leads to speed. And not just “speed” in the narrow sense of automating processes, but in the broader sense of speed of operations.
Huge, diverse service provider environments will never be exactly like homogeneous hyperscale clouds. Fortunately, there are now open, intent-driven automation solutions built specifically for service providers. They let you take the most innovative hyperscale operational concepts and apply them to the unique requirements of telecommunications DC networks. As long as you stay focused on what matters when building out next-generation DCs, you can bring the cloud revolution to your business.