How much does your ISP shape your downloads?

A global ISP tracker called “The Network is Aware” shows the processed data of a test used to detect whether an ISP is throttling downloads that use a particular protocol.

According to the website, which is hosted at Syracuse University and overseen by Professor Milton Mueller and his team, the project specifically seeks to detect deep packet inspection (DPI).

It is able to do this thanks to the Glasnost test, an application deployed by Measurement Lab (MLab).

“Thanks to an initiative known as the Measurement Lab (MLab), supported by Google, the New America Foundation and the PlanetLab Consortium, the Glasnost test allows end users all over the world to test whether their ISP is blocking or throttling BitTorrent and other protocols,” states The Network is Aware website.

“The results are stored and made available to researchers.”

The website allows you to filter results by country, and South Africa is among the countries listed.

Only Telkom and iBurst made the list and neither operator has a convincing number of valid tests next to their name.

However, one hopes that more users might start using the test so that over time the list will become more comprehensive.

We conducted our own tests in the morning between 08h30 and 10h00 on a 4Mbps ADSL connection using an MWeb uncapped account, a Telkom uncapped account, and a capped WebAfrica Titan DSL account.

Most of the time the Glasnost test warned us that “the measurement data is too noisy to detect whether your ISP rate limits your download traffic.” This was despite attempts to minimise the amount of Internet traffic generated on the network.

The test with which we had the most success was the Telkom HTTP benchmark, which ran on ports 38929 and 44023.

According to Glasnost, there were no indications that Telkom limits HTTP downloads, though the maximum speed achieved was 2,250kbps.

Looking across all our tests, the best speeds we achieved were between 2,048kbps (2Mbps) and 2,697kbps. If that is considered the “expected speed” for the purposes of this exercise, then our tests yielded the following results:

• MWeb shaped torrent downloads heavily during the day, but not uploads.
• Telkom’s (far more expensive) uncapped account did not seem to shape BitTorrent downloads.
• WebAfrica also shaped BitTorrent downloads into the ground, showing speeds slower than that achieved on upload tests.
• Uploads usually ran at around 400kbps, close to what is expected from the type of ADSL connection we tested on.
• The accounts performed similarly on HTTP download tests.

Since tests take around 8 minutes each to run and need you to keep your connection as quiet as possible, the performance of NNTP and Flash video on various connection was left for a future exercise.