[mcs37]

Check this out:

http://www.wired.com/news/technology...,59111,00.html

Superdrive your broadband connection. I wonder if the current NICs in the PowerBooks can support the Fast TCP protocol, or would they require additional hardware? Either way I hope this technology works its way through academia and into the real world soon! I'm tired of the pathetic "broad" band they sell to us. Broadband should be at least 10 Mbps but more likely in the 100 - 1000 range. That's what real broadband is.

[BkueKanoodle]

Originally posted by mcs37:
Check this out:

http://www.wired.com/news/technology...,59111,00.html

Superdrive your broadband connection. I wonder if the current NICs in the PowerBooks can support the Fast TCP protocol, or would they require additional hardware? Either way I hope this technology works its way through academia and into the real world soon! I'm tired of the pathetic "broad" band they sell to us. Broadband should be at least 10 Mbps but more likely in the 100 - 1000 range. That's what real broadband is.

Yes it would work on your hardware, seeing as the OSI model allows for independence between the communication protocols and hardware. All this changes are some of the "rules on conduct" for tcp packets, it has absolutely nothing to to with the physical connection.

However, by the time you actually see this work its way into the real world, whatever your currently using will be obsolete, so it won't mattter

[mosch]

Right now the most common scheme to maximize bandwidth utilization is TCP Reno, which essentially works by trying to send things as fast as possible, and then backing off when it is determined that the send is proceeding too quickly. This is implemented in OS X, but it is not turned on by default, probably due to the fact that some of these congestion avoidance protocols don't play nicely with each other, and you can end up in a situation where two different mechanisms to maximize bandwidth usage combine and effectively destroy the throughput rate.

TCP Vegas is the followup to TCP Reno, and it's generally agreed that Vegas provides for greater efficiency and fairness than Reno, which in turn provides for greater efficiency than TCP without a congestion avoidance mechanism.

Fast TCP appears to be a slight modification of TCP Vegas, but it's not groundbreaking, and it will not allow for 6,000 times more traffic, except perhaps in some sort of very contrived laboratory situation. It might be useful to eek out a free 10 or 20% performance increase at the server side, but that's it.

[cambro]

Originally posted by mosch:
It might be useful to eek out a free 10 or 20% performance increase at the server side, but that's it.

True, but 10-20% is definitely nothing to sneeze at...especially when it requires no real change to the system beyond how it handles packets.

[mcs37]

Originally posted by cambro:
True, but 10-20% is definitely nothing to sneeze at...especially when it requires no real change to the system beyond how it handles packets.

Indeed, a 10-20% increase is certainly enough to get my attention. I don't know if the professor is entirely overoptimistic citing the 6,000 time "bandwidth" increase, but if we saw increases up to 100 times better, that's still pretty much good enough to get video on demand (although a 1,000 times increase would be preferable).

If MS and Disney are specifically listening to this Caltech professor, I doubt he's off by too much.

[macintologist]

Originally posted by mcs37:
Indeed, a 10-20% increase is certainly enough to get my attention. I don't know if the professor is entirely overoptimistic citing the 6,000 time "bandwidth" increase, but if we saw increases up to 100 times better, that's still pretty much good enough to get video on demand (although a 1,000 times increase would be preferable).

If MS and Disney are specifically listening to this Caltech professor, I doubt he's off by too much.

Fast TCP: Available only for Windows!