BOLTHOUSE%MCOPN1@eg.ti.COM (God is in the details -- Mies van der Rohe) (03/16/88)
I have just started receiving info-vax again, so I'm not completely aware of the context of this discussion. I have had some experience in this arena, however, and would gladly give everyone a memory dump if they'd care to read it. We at TI have a *large*, multi-vendor Ethernet. We run TCP/IP, DECnet, SDLC, XNS, Chaosnet, and other protocols on the same "cable", which is actually a couple dozen plants linked with MAC level bridges. We have "a few" VAXes, as well as IBM machines (30xx running MVS and 43xx running VM). Our main information services division is very Blue in its' orientation (I work for the defense electronics division only). To get from VAX to MVS, there are several possible solutions (I am presuming VMS in this discussion, since that's mostly what we are running...but the TCP/IP portions of what follows would certainly be applicable to UN*X-based systems). The best one to date is the DECnet/SNA Gateway. It provides a great deal of SNA functionality in an easily-managed package. Don't buy the current hardware; wait until the next generation (long-rumored) arrives. If you have a large network with lots of broadcast packets, the current hardware won't work very well. It also doesn't give you very fast transfer speeds (45 kbits/sec using DTF, 9.6 kbits/sec using RJE). Of course, if you use DTF software over the DECnet/SNA Gateway, that is effectively network to point, since you can get from DECnet to a single MVS system. Be aware also that DTF uses a server *and* client software; you must have the client software on the originating system. The server softare isn't cheap either. IBM would like you to buy a box from Interlink to get to a 9370, which then gets you to MVS (at least I haven't heard of an Interlink solution for MVS -- feel free to correct me if I'm wrong). There are also TCP/IP solutions from FIBRONIX and A.C.C. which might work. We have not tested these under MVS at this time. INTEL has a channel-connect controller (FASTPATH) which can be made to use TCP/IP as well (I believe the software is by A.C.C.); we have had good success with the controller, but have no experience with the software. To get from VAX to IBM/VM, there are really no possibilities to date using the DECnet/SNA Gateway. You can do 3270 terminal emulation and 3270 datastream work on a VM machine, and the MAIL stuff apparently works (we don't have it here), but DTF in particular isn't yet supported on VM. I don't know if it would work using VM/XA with an MVS guest; I suspect it might work but it might also be unsupported by DIGITAL. IBM has a channel-attached box, the IBM 8232, which talks TCP/IP with software running on a VM host. It is a very competent product, composed of a PC/AT with a beefed-up motherboard and a channel interface. We send 70 Kbytes/sec (oops...replace 'send' above with 'sent') to it before we noticed any degradation in FTP throughput. If IBM has made this equipment and software available for MVS machines, or if IBM has any plans to do so, I haven't heard about it. Full-screen TELNET is an extra-cost option if you go this route. FIBRONIX also has a solution for IBM/VM. It is supported implicitly by ComputerVision, which has software that assumes its' presence. (We have "a few" CV machines as well, hence our interest.) I believe that as a standalone product, it is not as competent as IBM's offering, but it *is* less expensive to acquire. Service will be more problematic, however. There is also a point to point solution by FLEXLINK using the FASTPATH controller from INTEL which gets you from a VAX/VMS system to a VM system, but since Ethernet was specified, I won't give you the details. Suffice it to say that I wouldn't buy it now that I've seen the IBM 8232. FLEXLINK may also have an MVS solution, but if they do, I'm not aware of it. If you were to contact IBM, they'd certainly be able to give you more info. We have tested those configurations noted above using VAX/VMS, DECnet, CONVEX UNIX, SUN/OS, Ultrix, and TCP/IP from several vendors. Now for some disclaimers. First, the above is my opinion only, and not that of my current employer, Texas Instruments Defense Electronics Information Systems. Next, it is not the opinion of my next employer, IBM Corp. And finally, your experiences will undoubtedly differ from ours, given that you have a different network configuration, probably using some different vendors. David L. Bolthouse Texas Instruments Defense Electronics Information Systems VAX System Support McKinney, TX bolthouse%mcopn1@eg.ti.com (until 3/23/88) bolthouse%mcopn1@eg.ti.cs-net (alternative if your mail doesn't like '.com')