Wavelo Event-Driven Architecture (EDA)

At Wavelo’s core is a system-agnostic, event-driven architecture—a flexible and robust platform that can seamlessly translate API calls into real-time events. This enables operators to accelerate AI-ready modernization initiatives without the need for expensive, high-risk system replacements.

Frequently Asked Questions

Q: Is EDA a proven approach?
A: Yes. EDA has been used by multiple industries to support growth. Uber, AirBNB, Amazon have all successfully deployed EDA at scale.
Q: What are some of the advantages of using EDA for BSS/OSS?
A: The advantages include:
- Enables data to be streamed in real-time to multiple AI applications in parallel
- Automatic triggering of orchestrated processes based on events
- Cloud-native and massively scalable
- Co-existence with existing systems (translate API calls into events)
- Simple integration to consume information (no upstream impacts).
Q: Is Wavelo’s EDA just Kafka?
A: No. While we leverage Apache Kafka event-streaming capabilities, much of the value in Wavelo’s EDA is derived from what we have built around Kafka to make it easier to deploy in a Telco environment. Wavelo can generate, process and action messages. For example, Wavelo’s Notification Engine listens for event triggers and sends notifications to users, applications or systems. It can compose notification messages dynamically and also manages message delivery and retry logic. In short, we have built an EDA that is ‘industrialized’ for Telco.
Q: I can export data daily into a data lake - why would I need EDA?
A: Exporting to a data lake often means that data is batched with little context - it’s stale by the time it gets there. You also need to apply some form of common data model to be able to make more sense of the data that is being imported. This can be a lot of work. Plus the amount of resources required to ingest data from multiple systems every day and then map it to a new model - it can be quite inefficient. Wavelo’s real-time architecture allows you to map data on an event basis, which is considerably more efficient. It also means that current data can be served to LLMs and AI applications in real-time, on demand, with context.
Q: We have invested heavily in TMF APIs - do I have to throw that all away now?
A: Absolutely not. In fact, if you have TMF-complaint APIs already, we can shorten deployment times for you as we have already built a set of ‘listeners’ for the most common TMF APIs in use today. We continue to work with TMF to identify best practices for leveraging real-time data streaming in modern telco environments.
Q: What if I have multiple AI-based applications that require streamed event data?
A: We do not limit the number of applications that consume event data. Each system can subscribe to just the events that are relevant, and ignore anything that isn’t. If multiple systems all need to process the same event (perhaps for different purposes), they will all receive it at the same time, without the need for each of them to have a separate point to point API integration.
Q: I heard that EDA is not suited to telecom as it's too 'fire and forget'. The subscription approach is not deterministic enough in terms of message delivery.
A: At Boost for example (and as part of our offering), we have implemented Temporal, which directly addresses this objection. It provides a system of record for workflows, guarantees consistency across services, and delivers observability and replay capability. It also enables us to accelerate delivery by eliminating duplicate workflow logic across systems. Teams no longer build custom retry, compensation, or transaction-tracking code. Instead, we use Temporal to provide a unified orchestration layer so that developers can focus more on business functionality. For example, it allows operators to pause and replay workflows deterministically, ensuring continuity during outages or cyberattacks.

“Telecom's AI transformation has been hampered by a fundamental infrastructure challenge. Decades-old BSS/OSS systems were built for operational efficiency, not data accessibility, As CSPs urgently seek ways to scale their AI initiatives, the timing is ideal for solutions that can unlock this trapped data without requiring costly system overhauls or disrupting critical operations. Data extraction capabilities represent a critical enabler for operators who need to feed AI platforms with real-time data streams while minimizing operational risk.”

John Abraham, Principal Analyst

Appledore Research