DECnet High order bit issues

[John Galietta]

I am working on an old (problem has been occuring since 2009) issue for a client.  They have some VMS systems and for certain reasons must run DECnet.  They seem to be having intermittent file transfer issues between the VMS servers.  The issue occurs on the LAN and across the WAN.  At the moment I am focusing on the issue as it pertains to VMS servers on the same LAN (same VLAN!).

It seems that at one point Cisco and HP/DEC sent consultants in to examine the issue and it became a big finger pointing match.  What stuck with the customer was the determination of the DEC consultant that it was a network problem.  He blamed the Cisco switches and stated there was a problem with how they handled the high order bit.

Of course the client has no documentation from this engagement so I am left to do the detective work and I am not finding anything in my searches.  So my question is:

Has anyone heard of an issue with Cisco hardware and DECnet that has to do with the high order bit?

Thanks,

John

[Peter Paluch]

Hi John,

I have to state from the very beginning that I have no experience with the DECnet but somewhat of what you wrote intrigues me from a principial point of view.

So you are stating that some data transfer corruption is taking place in your DECnet network, and the DEC specialist complained about the Cisco switches "not handling the high order bit correctly". Do I get it correct?

Now, what kind of high order bit does the DEC specialist have in mind? The high order bit in every octet of transported PDUs, or just in a selected field (perhaps a header, a start of the payoload)? Anything more specific about the "high order bit" at all?

I am asking this because an Ethernet switch is, by its definition, a transparent device that does not modify frames crossing it, and it simply interprets the Ethernet frame headers to make switching decisions. I doubt that the Catalyst would actually modify the data in the Ethernet frames containing DECnet datagrams, thereby leading to data corruption.

If you have anything more specific about this issue please share that information with us!

Best regards,

Peter

[John Galietta]

Peter,

My experience with DECnet is limited and from the days before Cisco existed.  I haven't had much call (until this client) to dust off any of the old knowledge.  I fully agree with your assessment of the principal of the issue.  The switch should not modify any of the bits in a packet.  I believe that the issue may be a red herring tossed out by a frustrated "highpower DEC consultant" during the fingerpointing match between HP and Cisco.  I have been in touch with a retired DEC field engineer who told me that they used to label their trash cans "High Order Byte".

My research into this problem does point to an issue with the fact that the MAC address on a VMS system running DECnet is essentially locally administered.  That is, the DEC stack changed the MAC address to AA-00-04-00-xx-yy where xx-yy reflected the DECnet network address of the host.  However this appears to be a routing issue not a switching issue.  The reference is pasted below and found in this Cisco document:  http://www.cisco.com/en/US/customer/docs/ios/11_0/router/configuration/guide/cdecnet.html#wp613

Enable DECnet Phase IV Prime Routing

DECnet Phase IV requires that a MAC station address be constructed using  DECnet addressing conventions, with a standard high-order byte string  (AA-00-04-00) concatenated with the byte-swapped DECnet node address.  This can cause problems in configurations in which DECnet nodes need to  coexist with systems running protocols that have other MAC address  restrictions.

DECnet Phase IV Prime allows an arbitrary MAC address on the local-area  network (LAN). An address can be assigned globally (that is, assigned by  the IEEE), or it can be assigned locally by a system administrator.

To enable or disable DECnet Phase IV Prime, perform one of the following tasks as appropriate in global configuration mode:

Task	Command
Specify Phase IV Prime routing.	decnet [network-number] routing iv-prime decnet-address
Stop DECnet Phase IV or Phase IV Prime routing.	no decnet routing

My posting here is more of a fishing expedition to see if I can troll up anyone else who may have run into this same sort of thing.  The client is using GRE tunnels to pass the DECnet traffic between the sites with VMS servers.  The real issue is that they can't even consistently transfer files between two VMS servers on the same switch/VLAN.

Thanks for your reply,

John