Enterprises, webscale companies, and service providers rely on the internet to reach and serve their customers. However, the nature of many of today’s real-time applications and services demand faster speeds, microsecond response times, and high reliability. 5G services that promise to meet these demands are just starting to be deployed, and while 5G addresses much more than faster mobile access to the internet, it does not address one of the growing problems with the internet itself.
The internet provides global interconnection for thousands of networks and uses a process known as peering to distribute network reachability information, which enables routers to transport packets to their destination. The protocol used to achieve this is called the Border Gateway Protocol (BGP). But BGP – now in its fourth iteration – still lacks visibility into the capacity and utilization of network links and does not offer insights into traffic latency and application performance. Nor has it kept pace with the ever-increasing number of security threats occurring in the internet today.
This lack of visibility into traffic behavior in real time can lead to packet loss, unpredictable end-to-end performance, and poor reliability. In turn, this can impact customer experience, particularly for latency-sensitive video streaming, online gaming, and emerging applications such as augmented reality (AR), the internet of things (IoT) and industrial automation.
Further, BGP is complex to manage, and is frequently supported using highly manual processes in operations teams. This contributes to configuration errors that can inadvertently degrade performance, or cause networks to advertise incorrect routes, resulting in sites becoming unreachable; and potentially creating financial losses for effected companies. Although many of the problems of BGP are well understood, overcoming them in live networks today involves processes that are manual and reactive, and require constant monitoring and the engagement of skilled peering engineers.
To obtain first-hand information regarding the challenges, opportunities and evolution of network peering, ACG Research conducted interviews with CSP and webscale companies, including a globally active online gaming company.
This research showed that all these companies face several similar challenges, but their priorities and perspectives also differ in specific areas. Also, peering solutions are evolving toward use or more automated, intelligent processes. At the heart of these solutions is a transition toward insight-driven network automation, which enables dynamic network peering and eliminates many manual configuration processes and misconfiguration errors.
Enter insight-driven automation
Insight-driven automated peering provides a closed loop between each operator’s intent and the outcomes achievable in the systems they have deployed. It provides three key capabilities:
Real time visibility into traffic and network operating conditions, including fine-grained analytics on peering links, specific application flows, and establishing traffic baselines, trends and metrics
Traffic optimization, which identifies where traffic trends and metrics deviate from an operator’s intent and maps selected traffic flows to more appropriate paths
Network control, which directs traffic flows using standard protocols, such as Netconf/YANG, gRPC, REST, Openflow, BGP route injection and segment routing. It continuously monitors network state and resources in real time
ACG Research also identified three key capabilities that operator infrastructures require to enable insight-driven automated peering:
- Real-time analytics that leverage network and application telemetry
- Simplified and fully programmable control planes
- More capable data plane forwarding and highly instrumented peering fabrics
Automated peering acknowledges the dominance of BGP in existing networks while addressing its limitations with enhanced capabilities. The architectural framework used for automated peering can be combined with BGP Flow Specification to provide enhanced peering security with the ability to share DDoS mitigation information across networks.
There is also the human impact of automated peering to consider in its implementation. As Paul Parker-Johnson, Chief Analyst at ACG who carried out the research, comments, “To fully automate internet peering, we must build trust in the tools we are applying to the problem with the engineering and operations teams who are designing and running today’s peering implementations. Best-of-breed solutions incorporate dashboards and graphical displays that engage and enable human interaction with the tools, while developing trust along the way to achieving more fully automated and secure peering operations.”