Peak performance of DataStream 2000 service products
Imagine you go to buy a car called Ferrari 2000 at a very good price which you are told has an engine capable of going up to 192.0MPH. When you take it for a spin (on a racetrack of course, with the appropriate insurance), you can't get faster than 172.0MPH even when the roads are completely clear. You then go to have it serviced and explain to the dealer that you haven't been able to get the speed despite having a stig behind the wheel and you are then told that not all of these Ferrari 2000 cars can reach 192.0MPH and that this depends on how many of your neighbours have this particular car.. In fact if you're the only one in your neighbourhood, you'll have to cruise at 172.0MPH for a while. Sound familiar?
In a recent news article, we discussed Bulldog's PrimeTime 2000 service and how some users were never able to exceed 1720kbps and were told this was related to "contention" (the Internet equivalent to managed congestion). Since Bulldog's statement, we've discussed the issue with them over the last couple of days. They continue to believe that even with a peak rate of 1.7 Mbps overall throughput, the service is still good value for money. In fact, we couldn't agree more!--Even if it was sold as a 1.7 Mbps service, the overall throughput of the service represents exceptional value for money.
So why would the Ferrari owner be unhappy with a car that's excellent value for money and which whizzes past his friends Ford 500 at over three and a half times their speed? One would guess it's because he bought a car that could over 190MPH on an clear straight road.
What concerns us is the lack of clear information on the matter from the Bulldog support desk, in particular that the lower top speed was consistently attributed to contention. We feel Bulldog could be more open on the matter to its customers if they feel it's so insignificant.
Some could even question on where consumer law would stand on this issue since the service is being advertised as 'up to a blistering 2Mb' which suggests that users should see speeds up to 2 Mbps at some point in the early hours of the morning at least. Even comparing to a top speed of 480 Kbps (on 500 Kbps connection which is most common at present), speeds over 1.9 Mbps should be reachable. Would a court view it as reasonable that a user may have to wait some number of weeks before this is the case whilst more users subscribe to their service? If this is deemed reasonable, then what is a reasonable peak speed? 1.5 Mbps? 1 Mbps?
We should point out that the issue is not to do with the average speed on the Primetime 2000 service (which is what for example the ADSLguide speed test would show when averaged across all the users), but rather the top speed that even with a contended service people should reach at off peak times. Put simply, the PrimeTime 2000 service would appear to be capped at a peak of 1700kbps for the first users on each exchange, then as more customers join on that exchange this cap is removed as the virtual path is resized. On the positive side, this means the early adopters who cannot get more than 1700 Kbps will probably see far less contention than others, but on the negative side, you don't get what it "says on the box". It may therefore even be quite positive in some cases, but users are still being mislead as to the maximum speed attainable in those cases by around 10%.
The issue that even with a peak speed of 1700kbps the service is four times the speed we show 500kbps services as managing is a smokescreen in our view. Our speed test data shows the average speeds that the Top 10 providers are getting, the slowest of which last month managed an average of 433kbps, which if scaled to a 2000kbps headline speed would be 1732kbps, which looks good, but that is comparing an average speed to top speed. If one compares the peak speed of both providers and scale them accordingly, it would be 1920kbps versus 1700kbps on Bulldog. Another way to look at it is if Bulldog provisioned users on a 500kbps service with a 500kbps ATM backhaul their peak speed would be 434kbps. BT IPStream Home 500 actually requires 576kbps of ATM backhaul to manage the 480kbps peak TCP/IP throughput.
Bulldog was unwilling to discuss the precise details on how many users would be added to a 2Mbps (2000kbps) ATM virtual path before it was resized, i.e. the 1700kbps TCP/IP cap removed as they said this would reveal their planning rules and give competitors an unfair advantage. Though of course there are good financial reasons for this too, provisioning a 3Mbps virtual path rather than a 2Mbps one is very costly and with the low margins on the Datastream products, keeping costs down is very important to the product's viability.
So where do consumers stand? Well for those on the service the advice from Bulldog is that if you feel your peak speed is capped then contact support, who we hope will now log the matter and in time it will be dealt with. We would also like to point out that people experiencing slow speeds on 1Mbps and 2Mbps services need to ensure their system is tweaked for higher speed connections.
As the 1 Mbps and 2 Mbps products take off we look forward to having enough speed test results so we can publish some comparisons between the service providers. At that point it will be possible to see how the various Datastream providers manage contention on their service versus the traditional IPstream based services.
We have given Bulldog the opportunity to respond to the issues raised. They said:
"Bulldog is concerned that ADSLguide is tarnishing our image by making a big stir about a very small issue. Only a small subset of our 2Mbps users are temporarily affected and 'forced' to have a service with far superior throughput to what they paid for.
In terms of the legal issue of service description, it remains our commitment to give people as good a service as they pay for - or more. Our architecture for this handful of users is similar to the BT IPstream and Datastream Symmetric 2000 services. Like ours, these services are based on an ATM rate of 2048. They also have a data rate of 1855 (before TCP/IP and PPP headers), and do deliver a similar maximum throughput of 1.7Mbps. A number of service providers have already participated in trials of these services, which will be launched next month and 1.7Mbps bandwidth will become a de facto method of supply for services labelled 2000. Have no fear of 1.5Mbps.
In terms of clear communication from customer support, we would like to underscore that we have rapidly scaled our customer support organisation over the last month to support our subscriber growth. Many of our new staff have undergone on the job training and are now becoming aware of non-standard reasons for some service issues. Readers should realise and understand that contention is a frequent cause of people's inability to achieve top speeds. The staff have now been advised on how to communicate about this specific issue to customers. We apologise to anyone who received unhelpful or inaccurate support from us."
We would re-iterate that even at 1.7 Mbps speeds, the Bulldog service is excellent value for money, but we take issue with comments such as "a service with far superior throughput to what they paid for" as that implies that a low cost and very competitively priced service doesn't have to do what it says on the box.
Where any provider labels a service as providing upto a specific speed when it knows the service will not ever be able to peak close to this is worrying. We are concerned that this may be widespread in the emerging Datastream marketplace, with the result that early adopters may be dissappointed over their peak performance. ADSLGuide will be keeping an eye on this area and reporting back when we find similar occurrences. The Bulldog case was highlighted because support staff kept insisting to users that this is a contention problem which is not the case. Bulldog would be aware on placing orders on new exchanges that this will be an issue and to our knowledge, never advised end users of this.
Bulldog has this afternoon announced a series of network upgrades to address the concerns of users who are experiencing performance flucuations. Hopefully this will resolve the issues raised above to everyone's delight. Full Details in the Bulldog forum here.