Sorry, been a while since I stopped by. But my hosting company has managed to crash my website, so I'm wandering around feeling a bit lost right now.
Load tends to increase in a linear fashion, while capacity is added in big chunks, in a step-like function. So, depending on where you hit the capacity planning, in time, you will see better or worse congestion.
We monitor the network closely, but with around 6000 base stations, it becomes a numbers game, where is what priority?
Performance has continued to slide at Redhill, Durban. Basically a steady drift downhill. I understand the allocation of resources problem - I just think that informing users would help ease frustration - I can be quite understanding if I'm kept informed.
Your best bet is to report the congestion. These reports do end up in the right place and do see some kind of action, although it might not always seem so. You can also PM me.
Thanks for the offer. Where does one
officially report suspected congestion?
I find your observation on packet size interesting. I wonder if you're not seeing the effect of some sliding window in the data stream?
What throughput do you get when you do a FTP of a large file versus a small file? Does the throughput increase?
I think it's a case of you being overagressive on scaling transfer speed according to file size. Basically, I'm a webmaster. When I FTP a site, I get a list of transfer speeds. Next time I do a batch, I'll save the record for you. But it can range from 4 Kbps on teeny files to about 60Kbps on fairly large ones. There is definitely a speed to file size aspect ratio. you've commented on this before, I just think it's too aggressive at the bottom end.
When I'm downloading a backup, I'm smiling
Nothing faster I've come across when you're downloading 10 megs+!
The problem is a modern site can often have lots of tiny <1Kb graphics for buttons, etc. And this is where the speed scaling becomes an issue. The main stuff loads reasonably, the graphics come crawling in behind.
This site using vBulletin is a classic example. Try clearing your cache and then post a reply. See how long it takes for all the little buttons to load up. The difference between HSDPA and ADSL is very visible. Add congestion and things get fairly grim. Total pagesize is not much, but it's made of lots of little bits. Take a look at the source code in a post reply page and count the number of
img src=
To make matters worse, a lot of sites are setting aggressive no-cache values to stop caching problems for users of dynamic interactive sites. ISP's can be sooo naughty about caching. Overseas, bandwidth just isn't such a scarce resource but the ISP's seem to do it anyway
Sign up with Google
