<< | Thread Index | >> ]    [ << | Date Index | >> ]

To: "James Knott" <james.knott,AT,rogers,DOT,com>
Subject: Re: Slow file sharing performance
From: "Hans Steegers" <hsx,AT,dds,DOT,nl>
Date: Mon, 1 Sep 2003 14:46:45 +0200
Cc: <cipe-l,AT,inka,DOT,de>
Reply-to: "Hans Steegers" <steegers,AT,steegers,DOT,nl>

>I have tried cutting the MTU way back to 440 bytes.  The problem still
exists.
Where did you cut back the MTU? Making CIPE's MTU smaller doesn't change
anything if (for example) the sender sends 1500 byte packets, CIPE will have
to send two UDP packets to transport each of these packets over an ethernet
interface.

>Also, given that the external world sees only the encrypted packets,
>shouldn't things like MTU or ICMP affect every protocol and not just SMB
>or NFS?
No. The difference is that for file transfer using SMB or NFS the packets
are almost all full size. Packets for other protocols usually are small.
That is why I suspect that on one side the MTU/PMTU is handled incorrectly
(packet sent are too large...).

ADVICE: Use tcpdump / etherreal on both sides of the tunnel and dump the
packets before entering the tunnel as well as the udp packets when
transported to the other side. a difference between both sides will possibly
explain the asymetrical behaviour if packet size is involved.

If there are no differences, check again if ICMP trafic is not blocked (the
other possible explanation).

Try traceroute / tracert to see if there are differences in the routes from
each side that explain this asymetrical behaviour. If so, a misconfigured
router may be the cause. [ see the link on MTU on more info ].

** Asking again: how large is the difference?
If it is huge: timeouts because of ICMP protocol being (partly) blocked (?)
If it is moderate: extra fragmentation on one side because of an MTU issue
(?)

Hans Steegers

-----Original Message-----
From: James Knott <james.knott,AT,rogers,DOT,com>
To: Hans Steegers <steegers,AT,steegers,DOT,nl>
Cc: cipe-l,AT,inka,DOT,de <cipe-l,AT,inka,DOT,de>
Date: Monday, September 01, 2003 2:02 PM
Subject: Re: Slow file sharing performance

>Hans Steegers wrote:
>>>There are no firewall rules that apply to the VPN traffic.  The internet
>>>connection, that the VPN passes through doesnsn't see the SMB or NFS
>>>traffic, because they're encrypted in the tunnel packets.
>>
>> Of course, but a typo is easily made and overlooked causing filtering
>> between the LAN and CIPE interface...
>> Also: ICMP traffic can be involved for routing CIPE's UDP packets...
>>
>>
>>>CIPE automatically adjusts the MTU as required, but I have also tried
>>>smaller packets, without any improvement.
>>
>> The so called MTU problem is not CIPE's but other's having problems with
>> CIPE's 1442 or 1428 MTU.
>>
>
>I have tried cutting the MTU way back to 440 bytes.  The problem still
>exists.
>
>Also, given that the external world sees only the encrypted packets,
>shouldn't things like MTU or ICMP affect every protocol and not just SMB
>or NFS?
>
>
>--
>Message sent by the cipe-l,AT,inka,DOT,de mailing list.
>Unsubscribe: mail majordomo,AT,inka,DOT,de, "unsubscribe cipe-l" in body
>Other commands available with "help" in body to the same address.
>CIPE info and list archive:
<URL:http://sites.inka.de/~bigred/devel/cipe.html>


<< | Thread Index | >> ]    [ << | Date Index | >> ]