Re: SCP file transfer speed

[Stephan <stephanwib%googlemail.com@localhost>]

I´m actually having three boxes here at work:

One is a single-core Pentium 4 server @ 2,8 GHz on NetBSD 5.1 / FFSv1.
Two is an 8-core Xeon E5440 server @ 2,8 GHz on NetBSD 6.0 / FFSv2 WAPBL.
Three is a 4-core Xeon L5410 @ 2,33 GHz on NetBSD 7 BETA / FFSv2 WAPBL.

All of these have the same issue; transfer speeds are between 7,5 and
8,5 MB/s. That is why I suspected that this is a very basic issue.

Copying to /dev/null makes no significant difference.

2015-03-19 17:11 GMT+01:00 Greg Troxel <gdt%ir.bbn.com@localhost>:
>
> Stephan <stephanwib%googlemail.com@localhost> writes:
>
>> I performed a quick benchmark with netio and it showed up the best
>> possible speed on TCP.
>>
>> NETIO - Network Throughput Benchmark, Version 1.26
>> (C) 1997-2005 Kai Uwe Rommel
>>
>> TCP connection established.
>> Packet size  1k bytes:  11506 KByte/s Tx,  11116 KByte/s Rx.
>> Packet size  2k bytes:  11512 KByte/s Tx,  11469 KByte/s Rx.
>> Packet size  4k bytes:  11513 KByte/s Tx,  11469 KByte/s Rx.
>> Packet size  8k bytes:  11513 KByte/s Tx,  11100 KByte/s Rx.
>> Packet size 16k bytes:  11513 KByte/s Tx,  11469 KByte/s Rx.
>> Packet size 32k bytes:  11513 KByte/s Tx,  11470 KByte/s Rx.
>> Done.
>
> That looks good; 91.7 Mb/s and I dimly recall 93 as max theoretically
> possible with framing, IP, TCP headers.
>
>> UDP, however, does not work out of the box. It continuously displays
>> these errors:
>>
>> sendto(): No buffer space available
>
> Well, that's not really not working.  A modern cpu can call sendto()
> faster than the interface can send packets, and without any sort of flow
> control or congestion control, any buffer will be overwhelmed.  But as
> you say it's not relevant.
>
>> During an (uncompressed) scp copy, CPU usage on the receiver side is
>> largely below 10 % and disk utilization about 50% (which makes me
>> wonder because this disk can make about 100MB/s on sequential
>> write). I can copy to other Linux and Solaris boxes in this network
>> with about 10 - 11 MB/s from the NetBSD boxes - just receiving is
>> slow.
>
> Is it just one NetBSD box that is slow?  How many did you try?
> Do you have wapbl enabled?  Which fs?
>
> Try copying to /dev/null instead, to take the disk out of the experiment.
> I just copied to the disk (I had used /dev/null before) on one of my
> normal machines, and still got 10 MB/s.