General discussion

Locked

Transmission Control Protocol

By Dave_Email ·
I have heard it said that the TCP protocol becomes inefficient when used over high-bandwidth connections due to the window size being smaller than the amount of data transmitted in one cycle. Is there any truth in this? What about increasing packet size/sliding window size? I know that doing either or both of these things will require a larger buffer at the receiver end but that is in the hardware and the hardware should be designed to communicate over the maximum bandwidth that its interface supports... Anyone know the answer to this one?

This conversation is currently closed to new comments.

7 total posts (Page 1 of 1)  
| Thread display: Collapse - | Expand +

All Comments

Collapse -

Transmission Control Protocol

by Kevin Anderson In reply to Transmission Control Prot ...

The parameter you are after is found at

http://www.speedguide.net/Cable_modems/cable_reg_win2k.shtml

http://www.networkcomputing.com/1013/1013ws1.html

http://www.ncsa.uiuc.edu/People/vwelch/net_perf/tcp_windows.html

Collapse -

Transmission Control Protocol

by Dave_Email In reply to Transmission Control Prot ...

The question was auto-closed by TechRepublic

Collapse -

Transmission Control Protocol

by Kevin Anderson In reply to Transmission Control Prot ...

What happens is that you can send packets to another PC. The window sets a limit on how much data can be transmitted without the remote PC saying 'I got it, keep sending'. If your have a high latency in your network, your PC may send the first fewpackets, and then stop, and wait for the other PC to acknowledge that it recieved them. Then it's clear to send again.

If you increase the Send Window, you are allowed to send more before you hear anything back. That makes it less likely that you will have to wait for acknowledgement. Instead of sending for 1 sencond, and recieving an acknowledgement in 2 (wasting 1 second), you could send for 3, and still recieve an acknowledgement after 2, meaning you will never wait while data is 'in transit'. You will be cleared to send more, before you reach the limit of data that can be outstanding.

If latency is low, this won't be an issue. But on the internet, this CAN make a difference, particularly to sites on different continents.

I sent you the link, because there are several changes that should be made together. Speedguide.com (the first link) will have a utility that can change everything all at the same time for you.

Kev.

Collapse -

Transmission Control Protocol

by Dave_Email In reply to Transmission Control Prot ...

The question was auto-closed by TechRepublic

Collapse -

Transmission Control Protocol

by estebandelatorre In reply to Transmission Control Prot ...

yes, that's true.
It all relys on routers and communicating devices over the internet.
There are a lot of reasons and calculating on how big shold be a packet depending on the BW, the errors of a line, drops, re-trasnmmisions, overhead of a packet, ackowleges transmitted, etc, etc
These complex variables affects each other in each hop (router) across the internet.
The routers (every router) manage those variables according line status and BW. It will not be very usefull to tune these values, due to the fact that they are all unpredictable and variables second by second

Collapse -

Transmission Control Protocol

by Dave_Email In reply to Transmission Control Prot ...

The question was auto-closed by TechRepublic

Collapse -

Transmission Control Protocol

by Dave_Email In reply to Transmission Control Prot ...

This question was auto closed due to inactivity

Back to Networks Forum
7 total posts (Page 1 of 1)  

Related Discussions

Related Forums