10 Comments

1. 

   Sanjana Hattotuwa

   January 2, 2008, 10:40 am

   Hi Chanuka,

   Good stuff, this. Wonder if / when our TRC will follow suit.

   On a related note, Dialog Telekom in Sri Lanka were forced to change their Wimax media campaign because of communications I had with them on the gross disconnect (pardon the pun) between what was promised in their ads and what some customers, including I, actually experienced.

   See http://ict4peace.wordpress.com/2007/11/25/dialog-listening-to-customers/

   Best wishes for the New Year,

   Sanjana

2. 

   Ranga

   January 2, 2008, 12:32 pm

   I guess regulation does have a positive side.

   Another step would be to force operators to disclose any traffic shaping control they use because frankly the users don’t know where the problem is coming from, and from where they’re blocked.

   Most of the time, the p2p download speeds (Business connections in SL) reach higher speeds than typical http downloads. Which indicates that the problem may not lie in the physical layer but lies in fact in the network layer/application layer.

   Sanjana’s case is a typical physical layer problem that is supposedly inherent within WiMax. Therefore the operators are quick to respond, because the problem is easily traced back to them. And the integrity of the new Wimax technology, which they use as a flagship, is compromised.

   But there are countless Sri Lankans who pay for a 2Mbps connection but get download speeds of 300Kbps. Not knowing where the real problem is, they don’t know whether to blame the ISP, the application, the destination or the protocol in which the data is sent/received.

3. 

   drac

   January 2, 2008, 9:42 pm

   Ranga:
   By P2P, you mean BitTorrent or emule? Both download from distributed sources (closest possible/best transferring seeds and Kademlia respectively). This is similar to using a localized content delivery network, so yes – P2P will be faster than HTTP for most use cases without any shaping whatsoever. Comparing P2P vs HTTP is not a great indicator of the relative speed of your connection though.

   Therein lies my problem (pretty much the same as Ranga’s point).
   256kbps speeds to where? From where? Sustained? Burst? Which protocols? With server side caching? Uncached? Is the browser cache primed? At what time is this test taking place? (peak? offpeak? weekends?). Where is the server located? How many hops from server to client?

   Any of those variables or a few more that I probably forgot have a bearing on the download speeds. Does this document actually cover all those variables or is it a best-case? If it’s a best-case then it’s not much better than saying upto X kbps – it simply isn’t consistently measurable. Artificially boosting the speeds for a given speed test URL is easy for any ISP worth their salt – all they need is some transparent proxying (if they don’t have it already) and accesses will appear much much faster than they really are. This is how NZ ISPs (and possibly .au) operate.

   Having been in the market for decent business level internet connectivity, I know that the contention ratios on most international links owned by SL ISPs are horrendous (we’re talking around 50-1 at a minimum here). There is practically no redundancy.

   Rather than having the ISPs make some subjective quasi-promise that they cannot possibly deliver – I’d like to see somewhat more focused and specific requirements to reduce contention on international links, reduce prices or generally throw more hardware at the problem to reduce congestion. Not going to happen, I guess – but a guy can dream.

4. 

   Chanuka

   January 3, 2008, 8:17 am

   Drac,

   Thanks for your valuable input.

   Your alternative suggestions are to force ISPs to (a) reduce the contention ratio on international links and (b) throw more hardware (Need not be alternative; because these are complimentary)

   Agree with both, but how to make it? For example, Ofcom enforces contention ratios, but what guarantees ISPs follow that always? Do you suggest regulators should inspect operators’ network setup at frequent intervals?

   On the other hand, enforcing a floor for speed gives the consumers some ground to take the issue with the operator. May not be the perfect solution, but better than nothing.

   Best solution have so far seen is by IDA, Singapore where they define a RTT ceiling not to local loop, but till first entry point to US. (This is fair, because selection of the provider of international link is within ISPs purview.) The beauty of this is any consumer can simply ping to find out the ISP keeps the promise.

   In other words, the regulator employs an extremely effective group of inspectors free of charge – because every consumer becomes an inspector!

5. 

   Ranga

   January 3, 2008, 11:37 am

   Chanuka,

   What drac says is that ISPs can artificially boost speeds for the test URL So even if Singapore’s citizen inspectors ping and measure latency, it could be well that the ISPs are giving priority to ICMP packets travelling to that specific destination.

   Practically, measuring QoS in a absolute scale is therefore impossible.

   In my opinion, the inspection and certification method seen in other engineering fields (such as concrete testing/weld testing) is the best way to go about things.

   A designated authority can check the QoS of all the connection, without disclosing their methodology prior to the testing, and make public the results after the testing to the public. Since all connections are tested, the results can be comparative and not absolute. Thus fostering a healthy competition to give a better service.

   For people who need specific requirements, such as Drac, they can always request the authority for a certification of a connection before they pay for it. And according to the client’s requirement the certification can be issued over periods of time.

   Might sound far fetched or even bureaucratic, but I’m just thinking out aloud…

6. 

   chanuka

   January 3, 2008, 12:23 pm

   Ranga,

   That is not how I read it. (anyway, Drac can clarify)

   I have nothing against regulators checking QoS from the supply side, but practically might not be feasible. It is a massive operation and needs to carry out at frequent intervals. I am not sure the all the regulators are in a position of finding resources of that level. May be Ofcom, but not everybody.

   Without expecting the bureaucratic, resource starved government regulators to ensure the QoS, the citizens might prefer a simple methodology they can carry out themselves in spite of the minor bottlenecks.

   True, an ISP can manipulate the system but that will be easily detected by pinging another site.

   Finally, as I said above it is not a question of this and that. Both approaches can co-exist.

7. 

   drac

   January 3, 2008, 4:52 pm

   Chanuka,
   I shared Ranga’s expectation that the regulator could carry out spot checks for certain metrics but I think you’re right – that’s unlikely to be either practical or achievable. Having a massively staffed TRC behemoth terrifies me anyway.

   To be clear, I’d welcome any kind of guarantee on the quality of a domestic/low end business internet link by Sri Lankan ISPs. Any guarantee at all is better than nothing, as you rightly observe.

   I’m just unconvinced that focusing on a single metric that can be easily manipulated by the ISP is the best way to go. Ranga has already pointed out how ICMP traffic can be fixed (as I did earlier for a HTTP example). Prioritizing a single type of packet traffic across a link by an ISP is virtually undetectable (see the whole net neutrality fiasco for example).

   I also feel that ICMP isn’t the best choice since we would be measuring latency, not transfer. I think consumers who complain about these things tend to focus almost exclusively on transfer speeds. Obviously, even dialup can have magnificent latency – it doesn’t address the issue of being upsold a 2Mbps link only to get a 30kbps maximum transfer.

   Is automated assessment of ISP link quality (obviously, you’d need their co-operation) that difficult? I’m probably missing something here, but monitoring link QoS has been automated for decades. Can forcing ISPs to maintain and periodically hand over their link QoS statistics to the regulator possibly work? The regulator wouldn’t need to do anything except publish them. Concerned individuals and organizations would do the analysis. Come to think of it, the marketing departments of other ISPs would probably do their own analysis of competitors :) It would be just like publicly auditable financial records.

8. 

   Chanuka

   January 4, 2008, 10:21 am

   Drac,

   Fine. If everything works that way, we all can live happily ever after.

   But in this non-ideal world of ours things are not that simple and straightforward. I have no practical experience in telecom regulation, but having worked as a Bank Supervisor for 2 years I have an idea about the level of resources we need for effective financial regulation. Enough to say Bank Supervision was the largest department in Central Bank when I was there.

   Let me take this simple example. Take bread. What is the process of ensuring the quality of bread we buy? Does government place supervisors at every bakery or want them to declare information? No. They have made it possible for us to complain to Consumer Protection Authority if we are not satisfied.

   My suggestion is have a similar mechanism for broadband also but let consumers be educated in how to check the quality of their link. I think that is a worthwhile role every regulator can and should play.

9. 

   nitesh bariyar

   August 18, 2008, 3:59 pm

   i have taken aconnection of reliance and they told us that they will give us speed of 150kbps but they are not providing us even 100kbps how can i complain against them send me mail on nbariyar2indiatimes.com

10. 

    cristina aquinta

    September 2, 2008, 12:17 am

    hi, cool web site and good articles.
    The The first step is to learn how to find out what your Internet speed is.The Internet speed is determined by 2 components: the download rate and the upload rate of data which are usually measured in Kilobits per second.
    Then after you know the upload and download rate, you can use tools or tweaks to improve your internet speed, or take the decision to change your ISP.