There’s a growing view across the industry that OSS / BSS architectures need to change. At the same time open source big data analytics technologies are now being widely applied outside of big web companies and yet there is no coherent industry direction on how OSS and operations need to change in the presence of these new technologies. As a result, network operators are struggling to understand how to evolve their OSS and how to apply big data analytics and are looking for more open approaches
In this post we propose that big data analytics can be used to realise the next generation of operational analysis functions, analysing the data produced by these services and providing operational and business insight as fast as they can be provisioned.
Operation Support Systems (OSS) and Business Support Systems (BSS) cover a broad range of functions and have been defined by a number of industry efforts, e.g. TM Forum NGOSS and more recently Frameworx. A very simplified view of an OSS / BSS stack looks something like this:
The customer places an order for a service, which is exposed as a Customer Facing Service (CFS) by the BSS layer. The BSS layer deals with the management of the order; it understands the customer but has no knowledge of how to instantiate the service on the underlying infrastructure, so it hands off the request to the OSS layer, which exposes Resource Facing Services (RFS). The OSS layer deals with the instantiation of the service through provisioning and activation of the underlying infrastructure. Once the service is activated, the same stack deals with the management and monitoring of the service, and the abstraction of the resulting information up through the OSS / BSS layers, until it is represented back to the customer.
Drilling down on the OSS layer, its functions are often referred to with the – now very dated – acronym FCAPS, representing: Fault Management, Configuration Management, Assurance, Performance Management and Security Management. Huge industry efforts are being applied to virtualisation, automation, orchestration and control to make real-time network service provisioning possible, and open source software implementations supporting this space are maturing. These efforts supersede the ‘C’ in FCAPS; but what about the F_APS?
What about the F_APS?
In comparison, industry OSS architectures are not keeping pace with these developments; simply put, there are no definitive or de facto ways to monitor and analyse the data produced by these services as fast as they can be provisioned. The software technology landscape in this space has been transformed by the big data analytics movement and we believe that data analysis in these environments has become a big data problem, i.e. all OSS monitoring analysis applications can be addressed by performing a function against the complete OSS data set, e.g.:
- Fault management = ƒ(event data)
- Accounting =ƒ(metric data)
- Performance management =ƒ(metric data)
- Security analytics =ƒ(metric data, route data, flow data)
- Capacity management =ƒ(metric data, route data)
- Log search =ƒ(event data)
- Billing mediation =ƒ(event data, metric data)
Hence, we refer to these OSS monitoring and analysis functions more broadly as OSS Analytics.
Bifurcation of OSS
This ultimately leads to a bifurcation of the OSS / BSS stack, i.e. because the technologies that are used to analyse the data produced by services are fundamentally different than those used to orchestrate them, i.e. which results in something like this:
In this approach, orchestration/control and OSS analytics are related as loosely coupled systems, where the orchestration and control state provides the context required for the big data analytics applications to provide meaningful insight. The output from the analytics applications may be used to optimise the deployed services, closing the feedback loop to the orchestration and control functions. How then to realise these OSS analytics applications?
PNDA.io is an open source big data analytics platform which can form the basis of OSS Analytics for the next generation of NFV-enabled networks and services, which would look something like this:
- All data is aggregated and published to PNDA:
- logs/events, metrics and telemetry
- Across all domains; from infrastructure, from the services, and from the orchestration and control stack
- OSS and BSS functions are implemented as big data applications on PNDA or as applications taking a data stream from PNDA.
- The orchestration and control state provides the context required for the big data analytics applications to provide meaningful insight; we refer to this context as the Real-time Inventory
- The output from the analytics applications may be used to optimise the deployed services through feedback to the orchestration and control functions. This can be considered a realisation of Knowledge Defined Networking.
For more on applying PNDA to ETSI NFV see: ETSI NFV and Big Data Analytics with PNDA