Closing the Loop: Service Assurance and Topology

The key role of structural network and service visibility

To close the service assurance loop, it must be possible for the kinds of decisions and actions currently undertaken by network operations staff to instead be undertaken by software. The key inputs used by NOC staff to make decisions certainly include measurement and performance data – and modern analytics goes a long way to providing the kind of interpretation of this data that human experts provide. However, the other key knowledge domain available to operators is the structure of the network and services: the topology.

Astonishingly little progress has been made in automating the assembly of a comprehensive, cross-domain view of the structure of network suitable for machine consumption. 

Network inventories, which once promised to achieve such a view, failed on two fronts: they failed to develop reconciliation systems capable of reliably populating the Inventory model and they failed to find a way of making the model easy and cheap enough to change that it could keep up with rapidly evolving network designs and technologies. 

As an aside, it’s worth noting that the need to find a technically viable approach to cross-domain topology has more than one driver behind it. After a period of eschewing the need for a global view of structure altogether,  NFV reference architectures have evolved to accept that an “active inventory” and an associated up-to-date view of topology is required (AT&T’s oft-cited ECOMP framework calls these “Available and Active Inventory” and “Resource and Service Topology” – see http://about.att.com/content/dam/snrdocs/ecomp.pdf ),  and consensus is emerging that this should include legacy assets that are involved in the delivery of services on the “next generation” platform.

So, both in the interest of closed-loop service assurance and in the interest of the broader virtualization effort, this is a challenge that must be overcome. The good news for CSPs is that advances in data technology – including scalable event processing, stream computing and graph data representation – now put such a dynamic, comprehensive view of network and service topology within reach. Indeed, some operators have put such systems in production with great success.

What can we do right now?

The fact that extensive measurement infrastructure is widely deployed, effective analytics systems are widely available and proven solutions finally exist to the thorny problem of network and service visibility means that a form of “semi-automatic” service assurance, in which the work of operations staff in assuring services is vastly simplified (and thus cost significantly reduced), is possible right now.

Beyond immediate cost reduction, for CSPs, moving in this direction, is also an incremental step towards fully automatic, closed-loop service assurance and ultimately service orchestration. It’s an opportunity to discover and resolve some challenges early and use the knowledge to de-risk their planning for fully orchestrated NFV.

With the delays that have become apparent in achieving the promised benefits of NFV, opportunities to move incrementally (and bring some of these benefits forward) should be of great interest to CSPs.

Service assurance may well present just such an opportunity.