SUMMARY: dd block sizes for I/O over ssh tunnel across 100BaseT ethernet

From: Andrew J Caines <>
Date: Tue Nov 20 2001 - 18:07:49 EST
Fellow Manglers,

I am disappointed and at the same time slighly reassured to report that I
did not receive a single response with any answers, ideas or suggestions
relating to my question.

I did receive one question about an unrelated matter.

Since I have not had the opportunity to play with this myself, I don't
have anything useful to offer the list.

For posterity (and in case anyone would like to chip in), here's my
original question:

> I am writing a large amount of data over an ssh connection between two
> systems, client C and server S. On C I am sending STDOUT through a pipe to
> 'ssh $S "command file"', where "command" simply does a "redirect STDIN to
> file".
> What would be the optimal choice for "command" to maximise buffering and
> therefore minimise transfer time?
> A simple "cat > file" works, but if I understand cat, it buffers
> characters and is therefore a poor choice. Using "cat -u" turns off
> buffered output, but even if that helps, I suspect that input is the
> controling factor.
> dd seems like the obious tool for the job, but what value(s) should I use
> for input and/or output buffer size? I'd guess at n*BLOCKSIZE for obs, but
> what n? Should ibs be (n*)1500 for ethernet or 1500-(overhead for ssh and
> ...)?
> For the sake of simplicity, we may assume that S has lots of memory, bus,
> CPU and fast I/O, whereas C may not. The network is switched 100BaseT
> ethernet.

|  -Andrew J. Caines-   703-886-2689 |
| Unix Systems Engineer, WorldCom |
sunmanagers mailing list
Received on Tue Nov 20 23:07:41 2001

This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:42:28 EST