Re: Why I'm Skeptical of the FCC's Call for User Broadband Testing

[David Farber <dave () farber net>]

Begin forwarded message:

From: Richard Bennett <richard () bennett com>
Date: March 11, 2010 8:51:05 PM EST
To: dave () farber net
Cc: ip <ip () v2 listbox com>
Subject: Re: [IP] Re: Why I'm Skeptical of the FCC's Call for User Broadband Testing

I think it's safe to say that bursting or "burst boosting" if you will is the most common form of traffic shaping on 
today's Internet. It's great for people who are surfing the Web, as it boosts the performance of the most visible 
interactive aspect of the modern Internet.

The trouble with these tests is that they don't try to model any particular kinds of applications, and that should be 
the first thing the tester thinks of when designing the test: try a stream that looks like VoIP, try one that looks 
like a web page, try one that looks like YouTube, and then try one that looks like P2P. Instead of doing that, they do 
one generic test. It so happens that the smarter ISPs have optimized their bursting parameters for the generic case 
that Speedtest uses.

I ran both the M-Labs and Ooka tests from the FCC today, and got wildly different results: Ookla had me at 25 Mbps down 
and 2 ms of jitter, and M-Labs had me at  14.6 and 112. This disparity is to vast that it only says something about the 
tools, and not a thing about my connection speed and quality.

If you want a global view of Internet connection speeds, see the data from Speedtest drawn from users all over the 
world:

http://speedtest.net/global.php#0

It shows that users in the USA can easily buy a connection that's as fast as the average speed in the countries with 
the highest average speeds, so the fuss about national rankings is largely a question of what tier of service people 
choose to pay for.

RB

On 3/11/2010 5:35 PM, David Farber wrote:
> Begin forwarded message:
>
> From: Brett Glass<brett () lariat net>
> Date: March 11, 2010 4:55:39 PM EST
> To: dave () farber net, "ip"<ip () v2 listbox com>
> Subject: Re: [IP] Why I'm Skeptical of the FCC's Call for User Broadband Testing
>
> These tests will do little or nothing to determine a connection's bandwidth or quality.
>
> My network routes different types of traffic through different connections which are optimized for that type of 
> traffic. But the test doesn't "know" this; it tries to access random, uncacheable data through our cache and thus 
> gives results which are not typical of what users will experience. This hurts the outcome. So does the fact that our 
> network does traffic shaping with bursting (a technique which optimizes it for interactive activities while 
> deprioritizing unattended tasks such as long downloads) -- a feature which our users love.
>
> In short, the tests are "dumb" tests designed for "dumb" networks. My network uses innovative, "smart" techniques 
> which optimize the bang per buck for real life users but not for these artificial and unrealistic benchmarks.
>
> The tests also, for some reason I cannot determine, report wildly varying numbers for jitter. And they access only 
> certain destinations on the Net, including Google and "speedtest.net", thus biasing the results in favor of ISPs who 
> happen to be close, network-wise, to those sites.
>
> And they don't make sure that the user's connection is not congested by other traffic, nor that the user's computer 
> doesn't have a high processing load which hurts its responsiveness and hence the speed of the connection.
>
> In short, these tests appear designed to give the worst possible appraisal of an ISP's network, and to penalize an 
> ISP for implementing innovative technologies that optimize the use of scarce and expensive backbone bandwidth.
>
> --Brett Glass
>
>
>
>
>
> -------------------------------------------
> Archives: https://www.listbox.com/member/archive/247/=now
> RSS Feed: https://www.listbox.com/member/archive/rss/247/
> Powered by Listbox: http://www.listbox.com
>   

-- 
Richard Bennett
Research Fellow
Information Technology and Innovation Foundation
Washington, DC





-------------------------------------------
Archives: https://www.listbox.com/member/archive/247/=now
RSS Feed: https://www.listbox.com/member/archive/rss/247/
Powered by Listbox: http://www.listbox.com