Hi All,
I've got SSH/SSL (OpenSSH_4.3p2, OpenSSL 0.9.7l 28 Sep 2006)
running under AIX 5.3 (TL5, SP3). It is running on a 9112-265 with
2.5Gb RAM and runs nicely.

Logging on takes no time what so ever. Sub second.

The machine is connected to the Lan at 100Mb/s (verified from
entstat). It it connected via putty from an XP box, also at 100Mb/s.

File transfers using psftp or pscp (putty's secure ftp and cp
respectively) are running at about 20Kb/s!

The transfer via normal FTP is fine.

When the sftp/scp transfers are taking place, the CPU, network and
disk are not being overly loaded. CPU is <10%, disk is <20%.

This is driving me nuts!

Most of what I've read is about slow logons, which is not my
problem.

Anyone have any suggestions?

-Chris

On Tue, 07 Nov 2006 03:57:28 -0800, chrisg wrote:

> Hi All,
> I've got SSH/SSL (OpenSSH_4.3p2, OpenSSL 0.9.7l 28 Sep 2006)
> running under AIX 5.3 (TL5, SP3). It is running on a 9112-265 with
> 2.5Gb RAM and runs nicely.
>
> Logging on takes no time what so ever. Sub second.
>
> The machine is connected to the Lan at 100Mb/s (verified from
> entstat). It it connected via putty from an XP box, also at 100Mb/s.
>
> File transfers using psftp or pscp (putty's secure ftp and cp
> respectively) are running at about 20Kb/s!
>
> The transfer via normal FTP is fine.
>
> When the sftp/scp transfers are taking place, the CPU, network and
> disk are not being overly loaded. CPU is <10%, disk is <20%.
>
> This is driving me nuts!
>
> Most of what I've read is about slow logons, which is not my
> problem.
>
> Anyone have any suggestions?
>
> -Chris

What version of PuTTY are you using? Older versions had a slow transfer
problem. If you don't already have the latest (0.58) and try again.

chrisg@warpspeed.com.au wrote:
> Hi All,
> I've got SSH/SSL (OpenSSH_4.3p2, OpenSSL 0.9.7l 28 Sep 2006)
> running under AIX 5.3 (TL5, SP3). It is running on a 9112-265 with
> 2.5Gb RAM and runs nicely.
>
> Logging on takes no time what so ever. Sub second.
>
> The machine is connected to the Lan at 100Mb/s (verified from
> entstat). It it connected via putty from an XP box, also at 100Mb/s.
>
> File transfers using psftp or pscp (putty's secure ftp and cp
> respectively) are running at about 20Kb/s!
>
> The transfer via normal FTP is fine.
>
> When the sftp/scp transfers are taking place, the CPU, network and
> disk are not being overly loaded. CPU is <10%, disk is <20%.
>
> This is driving me nuts!
>
> Most of what I've read is about slow logons, which is not my
> problem.
>
> Anyone have any suggestions?
>
> -Chris
>

Do you get better performance with other scp or sftp clients? What
throughput do you get if you transfer to another aix server?

On Tue, 07 Nov 2006 03:57:28 -0800, chrisg wrote:

> Hi All,
> I've got SSH/SSL (OpenSSH_4.3p2, OpenSSL 0.9.7l 28 Sep 2006)
> running under AIX 5.3 (TL5, SP3). It is running on a 9112-265 with 2.5Gb
> RAM and runs nicely.
>
> Logging on takes no time what so ever. Sub second.
>
> The machine is connected to the Lan at 100Mb/s (verified from
> entstat).

What does entstat say about the duplex setting?

It's fairly common for managed switch ports to be fixed at 100Mb/s full
duplex. If this is the case the setting in AIX has to match. If AIX is set
to autonegotiate it'll end up at 100Mb/s half duplex. If it shows as half
duplex, try changing it to fixed 100Mb/s full.

Regards, Ian

F. Michael Orr <michael_orr25@yahoo.com> writes:
>On Tue, 07 Nov 2006 03:57:28 -0800, chrisg wrote:
>> File transfers using psftp or pscp (putty's secure ftp and cp
>> respectively) are running at about 20Kb/s!
>
>What version of PuTTY are you using? Older versions had a slow transfer
>problem. If you don't already have the latest (0.58) and try again.

There are further speedups in the post-0.58 development snapshots.

> What does entstat say about the duplex setting?
>
> It's fairly common for managed switch ports to be fixed at 100Mb/s full
> duplex. If this is the case the setting in AIX has to match. If AIX is set
> to autonegotiate it'll end up at 100Mb/s half duplex. If it shows as half
> duplex, try changing it to fixed 100Mb/s full.
>
> Regards, Ian

If this is the case and you changed it and it is still slow (So Ian has
a very good point!) then you colud also if you have a 1Gb card, change
the settings for the ent device:

Enable hardware TX TCP resegmentation yes

Enable hardware transmit and receive checksum yes

Change it to NO then ! (both off them) depending on your network
environment (check with the network guys)

Ian Northeast wrote:
> On Tue, 07 Nov 2006 03:57:28 -0800, chrisg wrote:
>
> > Hi All,
> > I've got SSH/SSL (OpenSSH_4.3p2, OpenSSL 0.9.7l 28 Sep 2006)
> > running under AIX 5.3 (TL5, SP3). It is running on a 9112-265 with 2.5Gb
> > RAM and runs nicely.
> >
> > Logging on takes no time what so ever. Sub second.
> >
> > The machine is connected to the Lan at 100Mb/s (verified from
> > entstat).
>
> What does entstat say about the duplex setting?
>
> It's fairly common for managed switch ports to be fixed at 100Mb/s full
> duplex. If this is the case the setting in AIX has to match. If AIX is set
> to autonegotiate it'll end up at 100Mb/s half duplex. If it shows as half
> duplex, try changing it to fixed 100Mb/s full.

Sorry, I thought that I'd said that (I didn't).

It is at full duplex.

# entstat -d en0
-------------------------------------------------------------
ETHERNET STATISTICS (en0) :
Device Type: IBM 10/100 Mbps Ethernet PCI Adapter (23100020)
Hardware Address: 00:02:55:76:1e:10
Elapsed Time: 0 days 21 hours 43 minutes 43 seconds

Transmit Statistics: Receive Statistics:
-------------------- -------------------
Packets: 488348 Packets: 693062
Bytes: 58246279 Bytes: 293541191
Interrupts: 305 Interrupts: 688495
Transmit Errors: 0 Receive Errors: 0
Packets Dropped: 4 Packets Dropped: 709
Bad Packets: 0
Max Packets on S/W Transmit Queue: 34
S/W Transmit Queue Overflow: 0
Current S/W+H/W Transmit Queue Length: 1

Broadcast Packets: 293 Broadcast Packets: 191719
Multicast Packets: 5 Multicast Packets: 0
No Carrier Sense: 0 CRC Errors: 709
DMA Underrun: 0 DMA Overrun: 0
Lost CTS Errors: 0 Alignment Errors: 0
Max Collision Errors: 0 No Resource Errors: 0
Late Collision Errors: 0 Receive Collision Errors:
0
Deferred: 0 Packet Too Short Errors:
0
SQE Test: 0 Packet Too Long Errors: 0
Timeout Errors: 0 Packets Discarded by
Adapter: 0
Single Collision Count: 0 Receiver Start Count: 0
Multiple Collision Count: 0
Current HW Transmit Queue Length: 1

General Statistics:
-------------------
No mbuf Errors: 0
Adapter Reset Count: 0
Adapter Data Rate: 200
Driver Flags: Up Broadcast Running
Simplex AlternateAddress 64BitSupport
PrivateSegment DataRateSet

IBM 10/100 Mbps Ethernet PCI Adapter Specific Statistics:
------------------------------------------------
Chip Version: 26
RJ45 Port Link Status : up
Media Speed Selected: 100 Mbps Full Duplex
Media Speed Running: 100 Mbps Full Duplex
Receive Pool Buffer Size: 384
No Receive Pool Buffer Errors: 0
Inter Packet Gap: 96
Adapter Restarts due to IOCTL commands: 4
Packets with Transmit collisions:
1 collisions: 0 6 collisions: 0 11 collisions: 0
2 collisions: 0 7 collisions: 0 12 collisions: 0
3 collisions: 0 8 collisions: 0 13 collisions: 0
4 collisions: 0 9 collisions: 0 14 collisions: 0
5 collisions: 0 10 collisions: 0 15 collisions: 0
Excessive deferral errors: 0x0
#

Jacob Nevins wrote:
> F. Michael Orr <michael_orr25@yahoo.com> writes:
> >On Tue, 07 Nov 2006 03:57:28 -0800, chrisg wrote:
> >> File transfers using psftp or pscp (putty's secure ftp and cp
> >> respectively) are running at about 20Kb/s!
> >
> >What version of PuTTY are you using? Older versions had a slow transfer
> >problem. If you don't already have the latest (0.58) and try again.
>
> There are further speedups in the post-0.58 development snapshots.

I was running:

Release 0.53b

Now running, 0.58, it peaked at 40Kb/sec and then dropped back to about
20Kb/sec.

Will look at other issues as well.

I tried/thought that I'd turned off all of NIS 2007 - perhaps I missed
a bit.

-Chris

Chuck wrote:
> chrisg@warpspeed.com.au wrote:
> > Hi All,
> > I've got SSH/SSL (OpenSSH_4.3p2, OpenSSL 0.9.7l 28 Sep 2006)
> > running under AIX 5.3 (TL5, SP3). It is running on a 9112-265 with
> > 2.5Gb RAM and runs nicely.
>
> Do you get better performance with other scp or sftp clients? What
> throughput do you get if you transfer to another aix server?

I guess that I'll try to set it up under AIX 5.1 as well.

Can you suggest another scp client?

A local loopback of a 4Mb file gave a transfer of 4.8Mb/Sec. :-)

-Chris

<chrisg@warpspeed.com.au> wrote in message
news:1162979186.169978.27700@m7g2000cwm.googlegrou ps.com...
> Sorry, I thought that I'd said that (I didn't).

... but are the pieces (i.e. the host ethernet card AND the switch port)
forced 100-full? i've learned you can't really trust what entstat says.. if
all is set auto or i'm forced 100full (but not forced at the switch), my
host thinks his connection is 100full but the switchport negotiates
100-half.. and then i get what you're seeing: some protocols seem okay,
others suck wind for transfers. i ususally do a quick verify by forcing the
host to 100-half and re-check the xfer rates: if they are better you're
still set to use auto somewhere and need to force speeds to make all right
again in the universe.

re-check entstat and look at the crc error count; has it increased
significantly?

what we've found at work: if 100Base-T, force both ends 100-Full. If GigE,
set both ends to auto (plus the other reccomendations for jumbo frames,
etc..).

-r

"No Body" <no@body.com> writes:

>.. but are the pieces (i.e. the host ethernet card AND the switch port)
>forced 100-full? i've learned you can't really trust what entstat says..

You can trust what entstat says, but it only shows what your host does.

To make it work both sides must be configured the same, i.e. both
sides must be set to autonegotiation or both sides must be set to
the same fixed speed and mode.

Mixing autonegotiation on one side with fixed speed on the other side
will not work because then there is no autonegotiation at all. Usually
that side falls back to some default setting (often 100-half).

In any case, when ftp transfers are fast and ssh transfers are not
then this can hardly be a problem with the duplex setting.

--
--
Michael van Elst
Internet: mlelstv@serpens.de
"A potential Snark may lurk in every tree."

No Body wrote:
> <chrisg@warpspeed.com.au> wrote in message
> news:1162979186.169978.27700@m7g2000cwm.googlegrou ps.com...
> > Sorry, I thought that I'd said that (I didn't).
>
> .. but are the pieces (i.e. the host ethernet card AND the switch port)
> forced 100-full? i've learned you can't really trust what entstat says.. if
> all is set auto or i'm forced 100full (but not forced at the switch), my
> host thinks his connection is 100full but the switchport negotiates
> 100-half.. and then i get what you're seeing: some protocols seem okay,
> others suck wind for transfers. i ususally do a quick verify by forcing the
> host to 100-half and re-check the xfer rates: if they are better you're
> still set to use auto somewhere and need to force speeds to make all right
> again in the universe.

I've just fixed it. I've been lied to!

You've nailed it perfectly, you've just described what I just found out
myself.

The Full Duplex led was not on on the switch.

entstat did say that it was running Full Duplex (i'd forced it too).

In the end, I let it auto negotiate, the switch is reporting full
duplex and I'm getting 2.5Mb/Sec transfers...
:-)))))))))))))))))))))))))))))))))))))))))))))))))) )))))))))))))))

So I'm happy.

Thanks for the tip. If I've not discovered it for myself, then I'd have
been truly stuck.

-Chris