Virsae Big Data Dive Lifts Lid on Gremlins Plaguing Voice Calls
Analysis of 257 million calls pinpoints lingering networking issues
Virsae Service Management (VSM), the cloud-native service management platform for unified communications (UC), is crunching a mountain of data to help IT managers root out remaining glitches in their voice communications systems.
In a first for the company, and possibly the entire unified communications category, Virsae has published the results of a data dive analyzing the quality of 257 million calls – just four weeks’ worth of data harvested from networks monitored by VSM.
Analyzing calls for latency, jitter, and delay, Virsae calculated a mean opinion score (MOS) for individual calls, with overall analysis showing the vast majority (99%) of calls were either good or satisfactory quality.
And while only a fraction (0.2%) of calls rated poorly, the percentage represents 400,000 calls, which in all likelihood detracted from the caller’s experience.
Of these poor-quality calls, analysis shows the majority were most likely undone by problems related to the endpoint configuration, including ethernet speed and duplex issues, misconfiguration of the QOS model, and mismatch between actual network topology and the requisite call server configuration.
Software releases, particularly in relation to endpoint DSP (Digital Signal Processor), were the biggest problem, with more than a third of releases out of date.
Virsae chief operating officer and company co-founder Ross Williams said analysis highlighted the potential for misdiagnosis when administrators weren’t equipped with the right service management tools.
“Administrators simply must have a window to underlying fundamentals before looking at the network – just as doctors must evaluate a range of symptoms before treating their patients.”
Williams said vendors like Microsoft had made it extremely simple to set up unified communications, but often users underestimated the planning required to keep voice functioning at peak performance when it competed with other data and users sharing the same network.
“UC systems that worked like a dream during a trial often wobble when call volumes spike and network users flit to YouTube on their lunch breaks,” he said.
And while call quality, or rather instances of poor call quality, were immediately apparent to callers, a more formal measure of quality helped network managers understand the performance they should expect, and diagnose weaknesses likely to degrade voice quality, said Williams.
On this front, the most reliable predictor of quality is based on the MOS Standard (Mean Opinion Score) – a transmission planning tool based on the ITU-T G.107 standard that calculates expected voice quality based on the experience of a typical telephone user under conversational conditions.
“In most cases, poor quality calls are the first indication that something isn’t right. Often the situation prompts managers to reach for historical records, which invariably point to a component on the network – a router perhaps. Only often it’s not the router,” said Williams. “A deeper dive often highlights a configuration issue – for example an endpoint sitting in the wrong VLAN, or QOS tagging that is not correctly prioritizing packets of data. So, while network monitoring flags a malfunctioning device, the root cause remains a mystery.”
Processing more than nine billion UC transactions every month, Virsae’s VSM cloud platform data analytics and predictive algorithms deliver a live picture of health for every component and its interrelationships within the UC environment.
View the Virsae EDUcast for an in-depth look at the causes of poor voice quality.