A great irony of virtualization is that it makes so many things much simpler, but brings huge complications. It makes operating, activating and using networks and services much easier, and more instantly gratifying, but it also makes them harder to monitor and manage.
That is why, as the telecoms industry continues to migrate from legacy physical networks to more virtualized architectures in the next few years, there’ll be a premium on assuring performance in an increasingly dynamic network and service environment.
That will be especially true in the emerging era of 5G. These mobile networks will have many more virtual elements, such as virtualized evolved packet cores (vEPC – a framework for providing converged voice and data), than previous generations of mobile networks. The vEPC concept has taken root in 4G networks, but it has been incorporated in 5G strategies from the start. This should allow operators to migrate quickly and painlessly to support the complex architectures and dense traffic scenarios associated with 5G, like the global evolution of the Internet of Things, and more dynamic and customized networks and applications.
Yet as vEPCs and other virtualized elements deliver network and application flexibility, as dynamic frameworks full of open source and off-the-shelf physical components, they could just as quickly threaten the ability of operators to manage quality.
The new monitoring challenge
Monitoring an increasingly virtualized network for performance is likely to become dramatically more complicated; it won’t be just about keeping an eye on devices that have a fixed, physical network presence. Hardware, chassis, fans and other components will still be important, but more network functions will be turned into applications running on virtual machines, which in turn run on servers. Virtual services will live in a mode of constant change as they are set up and torn down on-demand, and dynamically routed and controlled using OpenStack, SDN controllers and other orchestration systems.
Just looking at the OpenStack platform, there are several unique applications needing monitoring, more than 45 unique failure types and 50 different key performance indicators. Metrics must be pulled from virtual machines, and databases and logs must be checked. On top of that, physical hardware still needs to be monitored. More broadly, SDN controllers, virtual and physical network functions, the servers, storage components, power supplies and other aspects all need monitoring because they can impact service quality. It makes legacy network management using the Simple Network Management Protocol (SNMP) seem, well, simple by comparison.
In short, there is a long list of things to keep track of, and draw information and metrics from, as network managers look to create a topology map of a virtualized service at a given moment. And in the next moment, the reference points on that map could change.
How to keep up with constant change
For now, 5G networks are more talk than reality, and vEPCs have barely begun to see commercial deployment. The challenges posed by virtualization are just starting to appear. However, as virtualization advances to become a defining characteristic of 5G networks, any system or tool that has the goal of assuring network and service performance must be able to keep up with this tremendous fluidity of conditions, as well as be able to determine the impact of an issue on services.
Tackling network and service assurance in this emerging environment requires a full understanding of:
- virtual service chains;
- how all the open source and vendor-specific components are connected;
- all the scenarios that could impact these chains; and
- how to identify the root cause of any issues that occur.
As conditions in this sort of environment can’t be pre-calculated, network managers need tools enabled with graph database technology which can:
- provide real-time status and understanding of a given set of nodes and connections at any time to create a comprehensive map of a dynamic topology;
- offer automatic correlation to analyze collected data and determine the impact and root causes of problems; and
- support closed loop feedback to automatically optimize or fix performance issues or service failures.
Ultimately, virtualized 5G networks will need assurance tools that are carrier-grade. OpenStack and other open source network solutions will likely have their own monitoring tools, but they may not be carrier-grade.
Network and service assurance solutions must be suited to a hybrid world: thorough and capable of supporting carrier-grade service environments; savvy about the virtual service chain; and flexible enough for the new virtualized network. It’s the key to making virtualized networks reliable as well as dynamic and efficient.