Connections are – or at least were – a hot button issue for the newsgroup community. That and retention of course. One of the strongest selling points of a Usenet provider, so we’re often led to believe, is the number of connections provided. Recently, Usenet providers have been touting 20, 30 or even 50 connections. Is the extra price tag worth it?

Let’s talk about connections for a minute. Connections refer to the number of simultaneous processes that occur between a newsgroup client and the Usenet server. For simplicities sake, we’ll start off with one connection. Once an archive download is initiated, the newsreader will grab the article from the news server, complete the download, and request the next article. During that request, there’s a brief pause between article downloads, as the client prepares to receive data (watch our video on how connections work). This delay may not be much, but multiply that pause by many hundreds or thousands of articles in an archive, and we can see where multiple connections help out.

If we have two connections going, there’s a good chance we can overlap that delay with the additional data stream. But since these files are relatively small, the server and client are constantly starting and stopping data streams. Often times this prevents the article from downloading at full steam since the article isn’t large enough to take full advantage of the end user’s bandwidth. To resolve the start/stop nature of newsgroup downloading and the limits this places on download speed, we add more connections. The more data streams that are added, the better those delays are overlapped. Additionally, the inherent download speed limitation of such a small file is eliminated as more bandwidth per article is added. The bottom line - the more connections the faster our download will complete – to a point.

Traditional newsgroup wisdom preaches that multiple connections are a good thing. But like all good things, moderation is the key to success. So we put the ‘more is better’ philosophy to the test. Our testing method went as follows: a 50 megabyte dummy file was downloaded using Alt.Binz. The file was downloaded using 5, 10, 15, and 20 connections. Additionally, the file was downloaded on both a slow 740 kilobit DSL connection, and a much faster 8 megabit Cable connection. The "time" column displays the amount of time it took for the archive to finish downloading. The same computer (AMD Turion 64, 2 gigs of ram) was used for both tests. Now let’s see what happened.

Date: 2009-12-02 Source: http://www.slyck.com/news.php?story=1885