Static IP address?

[RogueKill]

For the last week since the move to the new IP range of 196.34.*.*, I have re-connected countless times and everytime I ended up with the same IP. Has anyone nocticed simmilar behavior with their connection?

 

[regardtv]

Quite right Rogue ... Same here ... I personally believe it has to do with the new way they are trying to "average out" or maintain the contention ratios ....

I take it you are not on a 128K package ? .. What size package are you on ?

R

************************************************************
The views expressed on this site are my own and NOT those of my employer.

 

[dorris]

Bandwidth pools are very easy to separate with netmasks rather than a huge lump of logins.
ANother crazy theory that might hold some ground to the situation, is that the collision domains are too large. With all users on the same subnet, 2000 people fighting on the same collision domain, means packets have to wait a long time to send (This could be the bottleneck currently being experienced) Splitting users into separate netmasks could alleviate the problem .

Come to think of it, this sounds like one of the most logical explanations I've heard/come up with, as we know (although some refuse to admit it) that its not the international pipe, as local is also crawling for those users with slow international, therefore the problem must be on the MyWIRELESS network.
This all holds, assuming mywireless works on the CSMA/CD technology, if any of those mywireless experts care to comment?
Could the collision domain be a problem affecting the towers?
I remember reading somewhere that each node has a capacity for 3000 users, so maybe this was taken care of??

then again the local peering point to IS, and the international peering point might both be the source of congestion.

Stab in the dark, Sentech hasn't done it for any reason, other than to issue static IP's to determine individual bandwidth, and throttle those users abusing the System! (We all know they are incapable of formulating this based on logins)

 

[TheRoDent]

Well, Sentech only has in the order of 1800 users countrywide, spread over numerous towers. On tower 126 Brixton, there are apparently only around 40 people. IPW's technology uses TDD (Time Division Duplexing) which guarantees everyone a fair slot, so I don't think it's the collision domain problem.

Frankly, I believe it's their peering, and possibly the Microwave links between towers that's causing the problems.

<center><h6> MyWireless <s>Hacks</s> Tweaks & Tech Info || Have you checked the fawking FAQ? <br /> <font color="red">Tired of bad Service? Want to compare speeds? We at least listen...</font id="red"></h6></center>

 

[regardtv]

Been doing some interesting testing with open tcp connections affecting bandwidth availability....

The more open tcp sessions - the less available throughput... anyone else notice this ?

Will write a few tcp packet/session transmitters and try to quantify ....

Easy way to test:

1) Start Getright/Flashget and download large(60MB) file from tucows.is.co.za
2) Once you see the connection stabilized - open eMule/EDonkey ...
3) bye bye bandwidth

* Note that running (2) I'm not doing ANY downloads etc etc .. It's incomming connecitons from other eMule clients.

R

************************************************************
The views expressed on this site are my own and NOT those of my employer.

 

[nonroker]

That's TCP/IP behaviour. If your upload pipe is filled, your downloads will start to suffer because the ACKs required to keep the data flowing struggles to get to the source.

By doing some traffic shaping and dedicating a small portion of your upload bandwidth to control packets you'll notice a big difference in download speeds while uploading

 

[RogueKill]

I am indeed on a 256K package.

Any idea what the upload bandwidth of a 256k package might be? 128K?
I have tried capping my upload to 10Kb/s however the moment I do that my download becomes non-existent.

 

[regardtv]

Actaully nonroker --- the CHANGE in behaviour vs past testing is what interests me ....

The influence of incomming connections is now far more of an issue than it was before ... From what ProASM has indicated (capping on local tower) I think that it has influenced things quite a bit

R

************************************************************
The views expressed on this site are my own and NOT those of my employer.

 

[nonroker]

Ofcourse it did, if you look back, the outgoing speeds was never properly capped. People on 128K getting 40KB/s uploads etc etc. The download speeds was never really affected by the uploads because of that.

 

[gripen]

It still depends on prevailing conditions. During full speed times, I see no difference with capping or uncapping my uploads. However when there are speed issues and speeds are capped &lt; package, it seems that capping uploads does increase overall performance. I.o.w ideally, it makes no difference since there isn't a "hard" cap on upload speed which makes TCP performance pretty good as compared to ISDN etc (if enough bandwidth is available)

 

[regardtv]

As an aside - I am currently doing shaping on the connection... but I believe the problem is elsewhere ... the service is now almost completely unusable for interactive use - it was bad before ( 200ms odd - but acceptable) ... now I'm sitting in excess of 400ms mmmm ... seperate issue but all connected.

************************************************************
The views expressed on this site are my own and NOT those of my employer.