HPN-SSH question

[Bjoern A. Zeeb]

On 16. Jan 2012, at 04:56 , Kevin Oberman wrote:

> On Sun, Jan 15, 2012 at 10:58 AM, Bjoern A. Zeeb <
> bzeeb-lists at lists.zabbadoz.net> wrote:
> 
>> On 15. Jan 2012, at 16:28 , Steven Hartland wrote:
>> 
>>> Yep we installed openssh-portable + HPN and experienced noticeably
>>> reduced transfer rates on high latency links, exactly the opposite
>>> as one would expect. I don't have the exact figures I'm afraid as
>>> we just went straight back to standard ssh.
>>> 
>>> I'll try and get some time to retest and provide some proper
>>> results.
>> 
>> Thanks. If you do please try the bundled version in 8-STABLE or 9 and
>> also gather the usual meta data (latency, packet loss, IPv4/v6, any
>> socket buffer tuning, ...).
>> 
> 
> And, if it happens with 9, any chance of a tcpdump capture of all of the
> headers for analysis? (No packet data needed.) We use the HPN version
> extensively on high latency links (often trans-oceanic) and have never seen
> that.  I'd find running tcptrace and generating time plots very useful for
> looking at this sort of performance issue.

Or using siftr.

/bz

-- 
Bjoern A. Zeeb                                 You have to have visions!
   It does not matter how good you are. It matters what good you do!