info-vax@ucbvax.ARPA (05/17/85)
From: Steve Aliff <Aliff@MIT-MULTICS.ARPA> Anyone care to relate their experiences with HYPERchannels and VAXen? Of great interest to me at the moment: Is there support for TCP/IP over HYPERchannels with either Unix or VMS?
info-vax@ucbvax.ARPA (05/17/85)
From: Ron Natalie <ron@BRL.ARPA> We use HYPERchannel here at BRL primarily because the Cyber main frame we have won't deal with any other type of network. Needless to say that we use TCP/IP. We have software that came with the Berkeley distributions and enhanced here to support it on our 4.2 BSD VAXEN and our extremely hacked V6 PDP-11s. In addition, our LSI-11 internet gateway also has an interface to the hyperchannel. The drivers are not trivial because the interface is wierd, but it didn't take two of our newer programmers that long to do it. The manual however is WRONG in a couple of places, best to copy the code from and existing implementation. Once we got the system put together it seems to work well and fast even though it is inordinately expensive for what it does. We'd really like to use the new Proteon fiber rings instead (and we will for things other than the cyber and the proposed supercomputers). No one wants to sell you a $3,000 lan interface for a $4 million dollar computer. The only problem we really have is that the adapters indicate a power failure sporadically but the comments in the code indicate that this is a problem with all the adapters. Provided you have the UNIBUS board strapped properly to pass these interrupts through to the host so that the software can reset things, these are fairly harmless. -Ron I heard a rumor that the NSC's hyperchannel specific protocol NETIX, is more inefficient in that limitted architecture that TCP is. Seems like they had to work real hard to do that.