GraphOps for the AI-enabled telco

A semantic layer for end-to-end, intelligent automation

This post is the second in a four-part series. The first article can be found here.

Background: The Telco Journey to the Digital Service Provider

Like other industries, Telecom is in the midst of a digital business transformation in the hopes of realizing CapEx and OpEx savings, supporting sustainability initiatives, and enabling transformative developer and customer experiences.

ETSI NFV was a critical first step in the journey to the Digital Service Provider. The NFV whitepaper laid out a vision that supported the transition from physical network appliances to virtualized network functions. The resulting standards initiative focused narrowly on defining interfaces for the foundational building blocks and their relationships. However, ETSI NFV did not attempt to describe how these newly disaggregated, virtualized and distributed network services would be practically managed at scale. A decade later, and industry standards bodies are still building interfaces from the bottom up without a clear over-arching vision of how everything comes together.

In the absence of a unified standard model for management of the networks, Telcos generally glue APIs together with imperative code and scripts, which brings the industry back to static, siloed RAN, Core and Transport domains. In effect, Telcos have replaced rigid infrastructure-bound networks with tightly-coupled software networks, which is why Lego™-like interoperability and agility remain elusive and NetOps automation and AI initiatives fall short of expectations.

This Infrastructure-as-Code (IaC) approach represents a half-measure in the softwarization of networks. Today, networks are essentially semi-automatic. There is no global visibility, end-to-end automation, or centralized policy management.

Despite a plethora of industry standards, these siloed automation stacks impede interoperability. Any variance in the version or implementation of standards can break interoperability as they are coded concretely. It doesn’t take a lot; any non-material, non-obvious, low-level detail that wasn’t specifically tested inevitably will be the source of a failure and since there is no domain model trouble-shooting is also manual.

“There is currently no standard metamodel for customer-facing services or standard way of modeling IT services. Leading-edge telcos say this is the largest service orchestration challenge they face.” Caroline Chappell

The lack of a harmonized standards-based Telecom model is a roadblock to AI-enabled telcos and autonomous networking. With the advent of generative AI, there is growing understanding that domain models, particularly graph-based models, are the key to connected and intelligent networks.

Why Graph?

Relational databases were made to record transactions at scale; however, they are hierarchical in nature. They are optimized for high-volumes of structured activities, but they are ill suited for highly-dynamic and unstructured activity.

Graphs are logical models of complex real-world relationships. They treat relationships as a first-class citizen. The design anticipates complex and evolving relationships as in a social-graph of family, friends and peers so they are inherently agile.

The Enterprise itself can be modeled as a graph of people, information and capabilities, bound by events and policies. Processes can be modeled as a set of loosely coupled tasks with relationships, bound by events and policies. This graph approach allows for data-driven processes that are based on real-time interaction context.

Similarly, all distributed systems, by their nature, are graphs — sets of connected nodes with relationships. Graphs are ideal for describing networks topologies and network services. Graphs can model the complex relationships between network functions, compute and network (i.e., a dependency graph), and can support varied and changing relationships in a way that static code and hierarchical information models cannot.

This is intuitive to most business people. The interrelationships between objects in a graph provides useful context for understanding events, supporting analytics and artificial intelligence. In short, graphs provide a programming language closer to human expression, allowing for a more natural human-computer interaction. This is a goal shared with Natural Language Programming and Generative AI, which is why graph knowledge bases are widely understood to be key to enabling both.

The fact that graphs are based on relationships makes it conducive to all forms of analytics and AI, which seek to identify patterns. Systems can expose graph domain models to algorithms so they can observe system activity and optimize decision-making. Graphs can also support AI directly by providing domain knowledge that informs its own processing.

The relationship between graphs and AI is symbiotic, but the graph domain model is the business asset. Algorithms are artifacts like scripts, there will always be many of them, but the domain model is something central to business operations. The graph provides coherence as a single source of truth, which is why “GraphOps” as in graph-driven NetDevOps automation is more apropos than simply AIOps — Graph is the enabler!

GraphOps

DevOps is often heralded as “shifting left” in the sense that developers, upstream in the process are given API access to infrastructure so they can programmatically manage what were once “Ops” tasks (i.e., Infrastructure as Code). While empowering at one level, it also puts an onus on developers who might not have the necessary expertise nor the time to take on this additional work.

GraphOps is an alternative approach. It’s about “shifting up” to a graph domain model, which supports declarative no-code design and zero-touch management. GraphOps is about simplifying and automating IT tasks and enabling intelligent and autonomous infrastructure applications by giving developers powerful abstractions. GraphOps connects teams and technologies under a unified abstraction to enable transparency, end-to-end automation, and consistent policy-based management — it’s a way to connect silos and end “swivel-chair” interoperability.

“The entire history of software engineering is that of the rise in levels of abstraction.” Grady Booch

Jinsung Choi, Deutsche Telekom SVP, Head of T-Labs, and Chair of the Board of the O-RAN Alliance, has had a series of LinkedIn posts putting forth forward-looking ideas for Telecom transformation, and in his most recent post he encapsulated the opportunity succinctly, “Abstraction allows us to break down the network into manageable layers, each focusing on a specific aspect. It also facilitates Automation and Orchestration: Tools and processes at different abstraction levels can be integrated and automated based on clear understanding of data flow and relationships.”

figure depicting high-level graph domain model abstracting complexity of Day 0, 1, 2 and beyond onboarding, composition, deployment and management of network services over Red Hat infrastructure management

Slide from upcoming GraphOps for the AI-enabled telco webinar

GraphOps is the next step towards the Digital Service provider. It requires a new class of graph-based automation systems that can support real-time, contextual and agile operations.

See GraphOps in action

GraphOps for the AI-enabled Telco

Secure Developer-centric Networking with CAMARA APIs

Intelligent Orchestration with Graphs: The Smart way to Coordinate SAP Services

Related Content:

Mandala Insights Whitepaper: Over-the-top orchestration for Telco Cloud

Analysys Mason Whitepaper: Metamodels: Flexible, extensible, adaptable approach to Network Function Onboarding and Lifecycle Management

Media coverage: Cloud, microservices, and data mess? Graph, ontology, and application fabric to the rescue

Graph conference talk: Graph-driven Orchestration

Contributed technical article: Graph Knowledge Base for Stateful Cloud-Native Applications

Bio: Dave Duggal, founder and CEO, EnterpriseWeb

Dave has spent his career building & turning around companies. He anticipated the challenges of increasingly fragmented IT estate and founded EnterpriseWeb to enable highly-automated and agile business operations. Dave is the inventor of 20 US patents on complex distributed systems. He is a regular speaker at industry conferences and an occasional blogger.

Register