Talking contention ratios at Telecoms World South Asia

Posted on October 9, 2008  /  3 Comments

At the end of a long day at Telecoms World South Asia in Dhaka, I presented some of the preliminary results of the Broadband QoSE work being done with IIT Madras. I talked about the finding that the bottleneck in Chennai and Colombo appeared to be the international segment and that the first results from the testing done in Dhaka suggested the same applied to Bangladesh, with the ISPs using satellite (versus undersea cable) were suffering very high latencies.

The CEO of a Pakistan ISP, Mr Wahaj us Siraj, said that the situation in Pakistan was very different, with plenty of capacity available on the undersea cables and low contention ratios (1:4) being used. Prices of international capacity had come down radically in recent times, he said, and now amount to only around 25 per cent of costs. I responded that we need to start testing in Pakistan soon, because this further illustrates the value of the AshokaTissa methodology, which allows the diagnosis of where problems exist which may vary from location to location. He was followed by Mr Undugodage, who was the driving force behind the introduction of ADSL to Sri Lanka, who said that they tried to give customers service equivalent to BT, using contention ratios of 1:15. This contrasted with the speaker from BSNL who blurted out that the standard contention ratio in his company was 1:20.

How is it that there can be so much variation in key parameter affecting the customer experience in three countries?

The slides that I used are here.


  1. @Rohan Samarajiva

    Just read the presentation… Do you happen to know how much the Sirius 256k broadband costs?

  2. Responding on behalf of Rohan.

    Sirius Broadband prices:

    We tested Xpress.

  3. Broadband QoSE testing is largely a virgin area and we learn so many new things as we move on.

    I recently learnt though sharing reduces the available bandwidth for each user, it doesn’t happen proportionately. For example, a contention ratio of 1:4 does not mean each user gets one fourth of bandwidth even in simultaneous use.

    Read somewhere BT uses a contention ratio of 1:20 while there are others who use ratios as high as 1:50. If BT uses a ratio of 1:20 that cannot be as bad as it sounds.

    As far as I know there is no way for a user to know the contention ratio(s) used by operator(s). All what users can do is to benchmark the ‘Service Experience’ from their end. (I may be wrong, please correct if so)

    We will learn more with each round of tests. Yesterday night I have been testing broadband speeds from Washington DC to 15 servers in different continents and the results were very interesting. Can you ever imagine that I get almost the same bandwidth from DC upto Vancouver, Rome, Kiev, Christchurch, KL, (and believe it or not) Mumbai, Lahore, Jakarta and Male? The only cities to which I could detect a bandwidth drop were Nairobi and Bangkok.(Sorry, not all cities are represented in No servers in Colombo, Delhi or Dhaka. But you may now try the Dhiraagu server in Male) The results are too different when tested from the other end.

    Showed the test results to two telecom regulatory experts I met today. They too are surprised. Can think of an explanation, but better not take guesses till we know the story in full.