Re: TCPIP performance for VMS
- From: jbriggs444@xxxxxxxxx
- Date: Thu, 23 Apr 2009 06:00:33 -0700 (PDT)
On Apr 23, 8:09 am, jbriggs...@xxxxxxxxx wrote:
On Apr 22, 7:28 pm, JF Mezei <jfmezei.spam...@xxxxxxxxxxxxx> wrote:
jbriggs...@xxxxxxxxx wrote:
However there's the interesting possibility that the PPPoE
encapsulation is reducing the available MTU
It isn't a possibility, it is a fact. The PPPoE kludge reduces MTU to
1492 (PPPoE header takes 8 btes minimum, and if you use MLPPP (multipe
links bonded into one), the headers are larger.
clear or ignored. A MAC using a MTU less than 1500 could be getting
its packets through intact while the DSL modem is forced to fragment
full sized 1500 byte frames being emitted by the VMS box.
It wouldn't be the DSL modem, it would bethe router (semantics...). The
modem always fragments the PPPoE kludge into ancient ATM packets (53
byte packet, 48 byte payload).
If the
speed testing site is detecting IP fragments, it could be falsely
concluding that PMTUD is not supported by the client
Thewww.speedguide.net:8080looksat the TCP options. And it says that
VMS doesn't have MTU discovery enabled. But the Mac has it. Both go
though the same router/link.
PMTUD isn't implemented in the TCP options. The DF bit is in the IP
header.
(semantics :-).
If you're still using Cisco router gear, there is a tweak you can try.
In interface configuration mode
ip tcp adjust-mss 1436 ! (or other selected value)
This causes the router to reach into the TCP SYN and SYN+ACK packets
and adjust the MSS advertised by the client to the server and the
server to the client. The upshot is that the two ends wind up sending
smaller packets.- Hide quoted text -
- Show quoted text -
Reading an earlier message from this morning, it appears that JF's NAT
router is already doing the same manipulations that "ip tcp adjust-
mss" would call for.
.
- References:
- TCPIP performance for VMS
- From: JF Mezei
- Re: TCPIP performance for VMS
- From: Jim
- Re: TCPIP performance for VMS
- From: jbriggs444
- Re: TCPIP performance for VMS
- From: JF Mezei
- Re: TCPIP performance for VMS
- From: jbriggs444
- TCPIP performance for VMS
- Prev by Date: Re: TCPIP performance for VMS
- Next by Date: Re: Should Oracle buy OpenVMS?
- Previous by thread: Re: TCPIP performance for VMS
- Next by thread: Re: TCPIP performance for VMS
- Index(es):
Relevant Pages
|