Service Management in the Brave, New, and Virtualized World
These are exciting times in the world of Network Functions Virtualization (NFV). NFV fever has truly gripped the market and NFV-washing is fully underway. However, there is real progress underneath and beyond the hype. Operators are beginning to seriously contemplate commercial deployments of NFV, and are devoting considerable effort to the important topic of NFV service management. Network operations teams at the service providers have been asking the difficult questions of how to manage all of the virtual stuff compared to the physical, custom built appliances they have operated in the past.
These questions include all of the traditional aspects of service management including service dashboards, SLA and performance management, fault management and capacity planning among others. The key to sorting all of this out is in recognizing that the underlying questions are not just about the new “virtual stuff” but rather about true end-to-end service management.
As part of our Ensemble OSA™ initiative, we at Overture have been thinking hard about these problems and will soon bring to bear new and exciting capabilities in response. In particular, we will soon announce new service intelligence tools to complement and work closely with our already launched Ensemble portfolio to address the NFV service management challenges we outline below.
The challenges for NFV service management can be categorized very broadly into data collection, storage, analysis and control. Let’s look at each of these in detail with particular attention to what is different in the virtualized world versus today’s model.
Data collection is a challenge, especially in a virtualized environment. One of the particular challenges with a virtualized implementation is that NFV-based services span a wide variety of physical network elements, virtual network elements, and cloud infrastructure. As a result data collection needs to address the collection of disparate data from heterogeneous data sources. Requiring each of these data sources to adhere to some kind of universal model would make the data collection plumbing very rigid and inflexible. Ideally, each source should be able to pump data into the service management engine in a native format that is ideally suited for that source. The hard work of analyzing and correlating that data, model fitting etc., is left to the applications and higher layer logic.
Then we come to the analysis of the collected data. The virtualized world brings a couple of different challenges:
- NFV services running in the cloud bring unique challenges because they are time-variant: they can be elastic, can auto-heal and can even migrate across the NFV infrastructure. As such, that the relationship between services and the underlying infrastructure is dynamic and not a static one. A runtime correlation between services and infrastructure is needed in order to track this agile and elastic environment.
- NFV infrastructure is shared by multiple services in a many-to-one relationship. If each of the services were to independently probe the infrastructure for health and performance reasons, this would lead to scalability and security issues. What is needed instead is a common infrastructure monitoring that can then be dynamically correlated to the service to provide a contextual identity for that piece of data.
Clearly, what is needed is the ability to collect and dynamically correlate diverse and disparate data in order to relate it to the NFV service(s).
Finally, we need to look at affecting control. As highlighted previously, cloud technologies bring the much-needed benefits of elasticity and healing. In order to realize these benefits, the data collection and analysis need to feed back into service orchestration to automate the responsive changes to the network and cloud resources. These controls alter the infrastructure and associated telemetry which feeds into data collection and a virtuous dynamic, adaptive cycle is established.
How do we address these issues of control and monitoring of a distributed and dynamic service infrastructure? The first step is to think holistically about creating and managing end-to-end services. We at Overture will soon be making some exciting announcements about how we will help provide solutions to these challenges.