Ahh, the hotly contested topic of traffic shaping, or now, traffic monitoring. Both terms mean roughly the same thing in the grand scheme of things, but not quite. This isn’t an article on the differences, this is an article on the best way to implement them.
We’ve all heard the story. AT&T in the US is going to start filtering traffic. Call it whatever you want, shaping, monitoring, filtering, giant glowing dildo, it all comes down to the same thing in the end, how to control content. In this day and age, it’s a lot harder to do than you think. We have sites like youtube, protocols like bittorrent, plus music streaming and other streaming content and you get a lot of bandwidth usage, and the backbones are crying because of how much it is. Again, not todays article.
Traffic shaping is a good thing!
Yes, you read that right
Traffic shaping is a good thing!
Well, with a few caveants. It must, MUST be done properly, and none of this bullshit way they do it now. There is very little room for error in doing this. Basically, internet users can be divided into 3 categories now. Heavy users, medium users, light users.
Heavy users are those that are always downloading something, playing games, uploading, sharing, HD video streaming users. They use the internet like it’s their personal toy, to hell with others.
Medium users, on the other hand, don’t do it quite so bad. they may play games sometimes and do some p2p sharing, and watch youtube, and use a fair amount of bandwidth, but nothing really compared to heavy users.
Light users only really use the internet to browse for information and check email. They rarely visit picture-heavy sites, or recieve emails larger than 100KB. I remember the days when that’d take over a minute to download to my computer.
Each plan also comes with a maximum speed. For example, my internet plan is a 10mbit/1mbit up/down cable connection. Theoretically, each second i can download 10mbits and upload 1mbits in perfect conditions. It’s fast, i get downloads around 1MB/s depending on the source. It was well worth it for me.
The problem comes in that if, for theoretical example, my block has a 100mbit source. Now, say on the block, sharing the 100mbit source there’s 25 customers, all on the same plan. math dictates that 25 customers would only get 4mbits each, a far cry from the 10mbit promised in our plan. The ISP oversells the source because they know that not everyone will be using the internet at the same time. Many connections are not constant (except streaming and p2p) and all traffic stops once said site/download is done. So if, by some fluke, 25 of us are all browsing at the same time and downloading stuff, chances are we will never see a slowdown because of that.
Now, lets add in the levels. Of those 25 customers, 5 are downloading large games that take hours to download, even at 10mbps. That’s 50mbps gone. 21 customers are now sharing 60mbit, but an additional 7 customers are watching youtube and streaming content, downloading some files, heavier internet, well say they’re also using about 45mbps. There goes 80mbit for 12 customers, leaving the other 13 sharing just 5mbps. If they want to download something, there’s no bandwidth. This is where traffic shaping could come in.
Traffic shaping could alleviate all that, but needs to be done well. To do this, you need to identify sources of really large files. Yes, bittorrent and other p2p networks qualify, but as do download sites like fileplanet and filefront, etc. This goes into a heavy group.
A medium group gets setup, where the streaming content is. Youtube, internet radio, what have you goes here.
A light group is the final. This includes picture files, email, regular old web pages, the like.
Now, There is no speed limits, just preset limits. Say, heavy group can’t go over 95mbit consumption no matter what time it is, leaving 5mbit for the rest. It also cannot go lower than 60mbit consumption (assuming there’s 50mbit of heavy stuff transversing the node), just to have both limits set.
Same thing with medium, and same thing with light traffic. This way during non-peak hours, the heavy users don’t get slowed down and bandwidth is utilised, rather than sitting there not doing anything. During peak hours, the heavy users still get a good chunk of bandwidth, but there’s holes for the rest, and it dynamically changes to accommidate the current traffic output, so there’s always 5mbit available for crisp, fast, web pages.
I dunno, maybe it’s just a dream, but it could work if built right, but you and I both know they won’t.
