Headlines News

Quality of VoIP in Africa – Tracing IP packets from the USA into Africa

08 December, 2013
ICT Africa News
8 December 2013

Africans, whether living in Africa or in the Diaspora, have explored many Voice-over-Internet Protocol (VoIP) services to beat the high cost of communicating with their loved ones. Viber, Skype, Vonage, MagicJack and Rebtel, are some of the services that are commonly used. Rebtel is a more recent and not well known service offering voice calls for as low as 4c a minute from the USA to Africa.

Unfortunately, for as long as the Internet Protocol (IP) Quality of Service (QoS) in many African countries is poor, the use of these services remain very unreliable. We have received a lot of feedback from a number of Africans in the Diaspora on the poor quality of VoIP calls from the USA to several African cities. Garbled words, parts of words missing, long delays between the time the caller speaks and the time the called party hears the words and echoes are some of the issues that were brought to our attention. Most of these problems are not experienced when POTS (or plain old telephone system) is used.

The Quality of VoIP calls from the USA to Zimbabwe
We made a number of test calls on the Vonage VoIP network to Econet Wireless Zimbabwe mobile phones on different days of the week. About fifty percent of the time, the communication was almost perfect without any noticeable glitches. The other fifty percent of the time the quality of the calls went from mediocre to very poor. In a few cases, the calls were dropped before a conversation could even begin. We were also able to replicate all the problems listed above from the different calls we made.

Most of the problems discussed here are caused by a number of network related attributes including latency, low bandwidth, jitter, and packet loss.

Latency (in ms) is the measure of the time delay between the time an IP packet is transmitted to the time it is detected at the receiver end. Latency depends on the distance between the communicating parties, the transmission medium used and the equipment in between. Geo synchronous earth satellite systems have the highest latency while optical fiber has the lowest latency. If the route between callers includes a geo synchronous satellite link, the latency will be high – at least 240ms just for the satellite portion of the route. The International Telecommunication Union (ITU) standard, ITU-T G.114 recommends a maximum of 150ms one- way latency for VoIP communications in order to achieve reasonable quality of service (QoS).

In an ideal situation, network latency is constant. In reality, latency changes as a function of time and the variation in latency is referred to as jitter. The higher the jitter, the worse is the QoS. We are not aware of any specific standard recommendation for jitter but a number of industry players have quoted 30ms as a guideline for the maximum jitter that can be tolerated after which VoIP and other sensitive applications, like gaming, are impacted.

Packet loss is the failure of one or more transmitted IP packets to arrive at their destination. The ITU G.729 standard recommends packet losses far less than 1 percent to avoid "audible errors".

Estimating VoIP impairments between the USA and Zimbabwe
In order to identity sources of VoIP impairments and estimate their magnitudes, one has to trace the path followed by IP packets between two points of interest. We have assumed the route between our offices in Elmira, New York, to an Econet wireless server in Harare, Zimbabwe – IP address - to be a typical path followed by our VoIP calls to Econet Wireless mobile phones.

We found Pingplotter to be the most ideal tool for tracing paths followed by IP packets and estimate their magnitudes.

We ran Pingplotter continuously for about 19 hours from 10:00pm on December 3 to 5:00pm on December 4 (Harare time) at 2.5s intervals. Figure 1 shows the path followed by IP packets, from the Time Warner network in Elmira, New York, through the Level 3 network in Chicago, through a number of other hops, through the Liquid Telecommunication network in London, Johannesburg and Harare before reaching the Econet Wireless server in Harare. There are a total of 25 hops traversed by the packets to Harare and back to Elmira.

 PingPlotter trace for USA to Zimbabwe VoIP route Pingplotter trace for Elmira, NY, to Harare, Zimbabwe

The black trace at the bottom of Figure 1 is the two-way latency between our office in Elmira and the Econet Wireless server in Harare. The latency ranges from about 300ms to more than 400ms. The higher latency will obviously cause a significant transmission delay if VoIp packets were to follow the route in question. The higher latency may be due to low bandwidth as a result of congestion and was contributed by the London-Johannesburg link of Liquid Telecommunications. The changes in latency constitute jitter which can severely degrade VoIP quality.

The red vertical lines at the bottom of Figure 1 denote 100% packet loss or time out and have severe consequences to VoIP and any other application for that matter. Every time there is a red line, a packet is either not reaching its destination or it is not getting back from the destination. If a VoIP call was placed at a time when there was 100% percent packet loss, it could lead to parts of words being cut or even result in a dropped call. There are three time windows, constituting about 25% of the testing time, with a high density of red lines which might be an indication of network downtime or severe network problems.

Due to the large number of data points and the time taken to generate the data, we chose not to calculate mean latency and jitter by calculating the average and variance from the entire data set. Instead, we transformed the data into the frequency domain by plotting a frequency histogram. This allowed us to throw out outliers due to errors and get a better understanding of the latency distribution.

Figure 2 is the frequency histogram of the latency data showing two dominant peaks with mean latency values of 301ms and 370ms and variances of 24ms and 169ms. The higher latencies and jitter were observed during the last 6 hours of the test and it is not clear whether this is a onetime event or recurring event. Further tests could shed some more light.

 PingPlotter trace histogram for USA to Zimbabwe VoIP route Pingplotter trace histogram for Elmira, NY, to Harare, Zimbabwe

From our test calls and PingPlotter test and analysis, there is more than 50% probability that one can complete a high quality VoIP call from the USA to an Econet Wireless mobile phone in Zimbabwe when packet losses are minimal, latency is about 300ms and jitter is low. However, rampant packet losses, high latency of 370ms and 169ms jitter will significantly degrade the quality of VoIP calls.

Operators in Africa, including Econet Wireless Zimbabwe should take advantage of the abundant submarine fiber optic capacity and significantly improve all aspects of terrestrial networks if better quality VoIP calls are to be a reality in Africa.

Please encourage our authors to put together more content like this by "liking" us below.



Posted by ICT Africa on 2013-12-08 23:21:05

You are right - it is obvious that Liquid Telecommunication is using fiber between London and Johannesburg; otherwise if satellite was used the latency could have been significantly higher. The author’s conclusion was to underscore that in addition to using submarine cables, every effort should be taken to keep impairments to a minimum. While for the most part, average latency was about 300ms – there was a time it went up to 370ms with very high variance (a measure of jitter) of more than 100ms. I am sure there is a way for Liquid Telecommunication to maintain the attributes to the lower values of latency and jitter measured in this experiment.

Posted by McTim on 2013-12-08 19:05:53

Your conclusion is not suported by your data. Yes, African operators should make use of abundant fiber capacity, but that is exactly what is going on in this ping! Econet has an upstream (seems like Liquid) which hauls the packets from LON to HRE via undersea cable. If it was via VSAT latency would be much higher, no?

Make a Comment | Email this to a friend

Tell A Friend

Name :
Your Email :
Friend Email :
Message :