Turning Siloed Data Lakes into Actionable, Real-Time Network Analytics

By: Naim Falandino

Cloud-based OTT applications and the Internet of Things (IoT) are putting demands on the network that are fundamentally changing not only the way it is architected, but also the way it is analyzed. Just as network silos and overlays are converging into a seamless dynamic mesh, operators need to join the isolated data lakes that currently exist — using big data to create a holistic view of the network. 

However, operators need to embrace that network analytics can no longer be reactive; they must be proactive and generate actionable insights that can accommodate and drive the dynamic nature of the new network.

Simpler Times

Not too long ago, Simple Network Management Protocol (SNMP) and telnet were the principal tools used by network operators to analyze the performance and health of their links. The main concern was being able to identify problems. SNMP was designed to work on sick links. It was lightweight and had a minimum of features so as not to overload what might already be a failing connection. If a link was fully functional, there was no need to analyze it in real-time. As a result, point monitoring — or polling every five minutes — was sufficient.

As network and data center architecture has evolved over the past 10 years, running a cluster with headroom has been replaced by running at maximum utilization. With the ability to scale up and orchestrate traffic at a moment’s notice, hyper-scale cloud companies — such as Google, Amazon and Facebook — have passed the limits of SNMP and need to know instantly what is happening and when changes need to be made to react to load. 

Instead of polling, they need asynchronous, event-driven updates that inform them of failures and their locations as they occur. And because their subscribers are intolerant of performance issues, they need to have a real-time view of the operational state of their entire network as well as the ability to precisely troubleshoot and remedy those issues.

Streaming Telemetry

Thankfully, the industry has recognized the limits of SNMP and is adopting streaming telemetry, with all router and switch vendors scheduled to support it by 2018. This will improve the performance over SNMP by up to two orders of magnitude and provide more granular and flexible data — a welcome change over the limited SNMP dataset, as streaming telemetry will provide cloud operators with more timely data.

State-of-the-art streaming telemetry will support thousands of network elements and services and collect data from hundreds of different data sources, as opposed to just existing SNMP Management Information Bases (MIBs).


Latest Updates

Subscribe to our YouTube Channel