DECnet+ copy very slow

[JKrucus]

We have a new DS15 Alpha VMS 7.3-2 server running DECnet+. When I copy a large file from any old server to this new one it runs fine. But when I copy from the new server to any old server it runs at 50 blocks per minute; it is also slow copying from an old server. It appears that files going to this new server run at full speed, but any file coming from this new server run at 1% speed. FTP has the same behavior. I must be missing something in the configuration.

[Volker Halle]

Hi,

does MC LANCP SHOW DEV/COUNTER show any errors at the LAN interface level ?

What's the auto-negotiation setting on the DS15 and the switch ? Ideally, both should be turned off and set to the same fixed speed and duplex mode to minimize problems.

Volker.

[JKrucus]

The switches and consoles are all set to 100 full duplex. I have 13 million Unrecognized multicast destination packet errors, and 900 alignment errors, 800 frame check errors and 200 receive data length errors.

[Volker Halle]

Hi,

do these error counters INCREASE, while you are copying a file and it seems slow ?

Are you really sure, that the switch is set to fixed speed and the same duplex setting as the NIC in the DS15 ?

There should be no frame-level errors between the switch port and the NIC in the DS15.

Volker.

[Wim Van den Wyngaert]

Check if you see the same values in mc lancp show dev/char (100 fd).

Do the error counters increase during the slow transfer ?

Is the performance of FTP as slow as the one in decnet ?

Wim

[JKrucus]

Yes the errors increase. I copied a file for one minute, only 200 blocks. Unrecognized multicast packet errors increased by 524, alignment errors the same, frame check errors +1 and receive data length errors +6. mcr lancp sho /char eia0 shows 100 full. The switch administrator says the port is set to 100 full, but I will have him verify this again.....this may take hours or days to verify. Thank, Jerry

[Wim Van den Wyngaert]

Enjoy some reading in the mean time.

http://support.3com.com/infodeli/tools/netmgt/tncsunix/product/091500/c11ploss.htm

Wim

[Volker Halle]

Jerry,

I'm pretty sure, that the switch port is set to auto-negotiation. This is the most 'popular' error found when receiving complaints about 'slow network file transfer' - especially in new installations (or after changing network equipment/connections).

Have your network person tell you the MAC address, which the switch reports on this port and verify that information with the current LAN address from LANCP> SHOW DEV/CHAR EIA0 - then you know you're talking about the correct switch port.

Volker.

[Ian Miller.]

and this one
http://www.mug.jhmi.edu/mirrors/InfoAlley/0896/08/physical.html

[Bojan Nemec]

Jerry,

You can experiment by changing the NIC speed and mode with:
$ mc lancp set device /autonegotiate /speed= /full_duplex

(see mc lancp help set device)
First I will try to set the NIC to auto which is the default on new switches and the only option on unmanaged switches.
When you get the right settings you can set the speed and mode in SRM console (my preffered) or with mc lancp set device /permanent or arrange with the switch administrator to set properly the switch port.

Bojan

[JKrucus]

The switch port was set to 100 full, no auto negotiate. I have the same issue. The copy is very slow and I am getting alignment errors and frame check errors on the target node.

[Jan van den Ende]

JKrucus,

if you have confirmed that the setting on the card and on the swich agree with one another, and you keep so many error and slow communication, then you prbably need to double-check cabling and connectors.
If you have any possibility, then replace the cable (if only temporarily for testing.

Remember, DecNET _IS_ a robust protocol, it will go to any length to keep your connection going, but the price for that in performance can be high.

hth

Proost.

Have one on me.

jpe

[Doug Phillips]

I experienced this same phenomenon once when doing a VAX to Alpha upgrade. I could log into the VAX (MV3100) and copy files to the Alpha (DS10) with very good performance. However, if I logged into the Alpha and tried to copy from the VAX the performance was terrible. The two systems were on the same ethernet segment so no switch was involved. I didn't try to solve the problem because my concern was in doing the upgrade but I did think about it for a bit.

I suspected that it had to do with the difference in the two system's performance and that Alpha was seeing the VAX slowness as a problem. I didn't explore it any further, though.

-Doug

[comarow]

Alignment errors are a serious problem.
Those are errors on the Data Level on NCP.

It is often and incorrect setting match, but it could be a bad card, cable, port, or switch. You can try another cable, port and see what changes. If you still have alignment errors either one of the settings is wrong or you have a hardware issue.

Some cards insist at being set to console level autonegotiate, while other's won't work with autonegotiate turned on. This may or may not be a problem. The network team is an expert on this area.


If not I hope you have a service contract.
If you have a valid serial number, feel free to email me at robert.comarow@hp.com
with contract or serial number information.

We can take it from there. You can also log a call from the itrc site if you wish.

Bob

[Jan van den Ende]

@ Bob Comarow:

Bob, for the Forums etiquette, if not
your own protection:
Please do NOT (ever) issue your email address in plain text.
Use any easily human decryptable description of such address, eg replacing . by "DOT" and @ by "AT", or insert recognisable characters to leave out, etc.
robertDOTcomarowAThpDOTcom works fine for humans, but will escape (most?) email miners.

Proost.

Have one on me.

jpe