From: SMTP%"RELAY-INFO-VAX@CRVAX.SRI.COM" 20-MAY-1994 19:51:07.88 To: EVERHART CC: Subj: Beware Upgrading Multi-Circuit Phase IV Networks To Phase V... Date: Thu, 19 May 1994 0:06:56 -0400 (EDT) From: "Clayton, Paul D." To: INFO-VAX@SRI.COM Cc: CLAYTON@radium.xrt.upenn.edu Message-Id: <940519000656.2180115b@radium.xrt.upenn.edu> Subject: Beware Upgrading Multi-Circuit Phase IV Networks To Phase V... First the background... The program I am currently working on is in the process of planning a massive upgrade effort to newer revisions of software such as OpenVMS 6.1 and DECnet/OSI 5.7 (Phase V). A coworker and I have been tasked to determine impacts of OpenVMS 6.1 and DECnet/OSI 5.7. No small challenge based on our operational suite of hardware and software. It would be absolutely correct to say that the most difficult portion has been, and will likely continue to be, understanding DECnet/OSI and how to move from a 'tweaked/special' Phase IV topology/configuration to a Phase V equivelant. Now the issue being raised... The most startling realization found so far can be whittled down to the following simple topology. ----------------- ------------ | VAX A | T1 based BRIDGED link (primary) | VAX B | | |-----------------------------------------------| | | | | | | | 56KB based BRIDGED link (backup) | | | |-----------------------------------------------| | --------------- ------------ With a Phase IV configuration of two L1 routers at opposite ends of the links, the primary link can be costed less then the backup and the result is that all DECnet traffic will use the T1 link as long as it is alive and working. In the event that the T1 link breaks, DECnet will automatically failover to the 56KB link and use it until the T1 link 'comes back'. This results in 'normal moment-to-moment' traffic being nice and quick over the T1, with a failsafe option to a slower (and cheaper) backup communications link with no intervention needed at all. This works great and there are all sorts of variations of this configuration that result in essentially the same thing. Under Phase V, there is a MAJOR wrinkle introduced. That being no apparent 'concept' of path costs on a per circuit basis when you configure a multi-circuit End-System under Phase V. The result of this when applied to the topology above, is that the Phase V End-Systems will determine that there are multiple paths between each other and send user data packets over BOTH links ALL the time! So HALF your traffic will zip along at T1 speed and the other half will end up on the 56KB doing snail speed. Overall network performance will be terrible. Needless to say, end users of such a configuration will be screaming from the rafters in very short order. There are a number of options from what has been learned so far. 1. Purchase expensive Brouters for BOTH ends and configure the Brouters with a higher cost on the 56KB line so that the T1 is used by default. But Brouters are not cheap and the topology may have been working very well like it is for years. Now due to one upgrade a major expense is incurred. Plus all the additional management tasks that are needed to get and keep Brouters working. 2. Set path splits to '1' on the End-Systems. The downside to this is that in this case, you could still end up with your '1' path being the 56KB line, even when the T1 is up and working fine. And for those of us with three Ethernet links between systems, where we want the bulk of traffic on one LAN segment only, and if that segment fails, then path split the load between the remaining two LANs, we are OUT of luck. 3. Physically disconnect one LAN segment so that only one path is found and therefore used. The downside here being that when the current path fails, human intervention is required to reconnect 'the other LAN'. This could be costly in both time and resources lost till corrective action is taken. I can only hope that DECnet/OSI Eng. will return the concept of path costs to End-Systems in VERY short order since it has been one of the hallmarks of DECnet up through Phase IV. Now I am left wondering what other nifty gotchas there are in Phase V land that we have yet to uncover. ;) Proceed at your own risk appears to be the motto of the day... pdc Paul D. Clayton Address - CLAYTON@RADIUM.XRT.UPENN.EDU (USA) Disclaimer: All thoughts and statements here are my own and NOT those of my employer, and are also not based on, or contain, restricted information.