UCCXKVB@OSUCC.BITNET (03/29/90)
AIX-L@PUCC.BITNET, IBM-MAIN@RICEVM1.BITNET When and if AIX/370 gets announced, Oklahoma State University, is planning to run it on an IBM 3090-200s. We have a few questions with AIX/370 with regard to connectivity. How do you attach an asynch. network to the AIX? For that matter, how do you connect any asynch term. to AIX. We have a feeling that we will have to dedicate several addresses off of a FEP to access the network. Is this possible? How can this be accomplised? We have an IBM 3705 and 3725 FEP for the 3090. We also can not afford to have a cluster of PS/2 servers with asynch ports hanging off the PS/2. We do not have the asynch option for our 3174 controllers. We do not expect purchasing the asynch option. We would also like to do AIX/370 from a 3270 terminal. Is this possible in full screen mode or is only available in line mode. The IBM FTN last November discussing AIX/370 and TCF mentioned that doing vi on the 3090 was very compute intensive since the 370 architecture does not handle UNIX full screen applications very well. How much of a resource drain are such UNIX full screen applications? How much of a resource drain is AIX in general (just to have it up and running. Other comments are also appreciated.) (This vi/compute intensive issue came from a user question at the end of the broadcast.) Thanks in advance. Konrad Brandemuhl UCCXKVB@OSUCC University Computer Center 113 Math Sciences Bldg Oklahoma State University Stillwater, OK 74078-0606
ALTON@DUKEVM.BITNET (Alton Brantley) (03/29/90)
On Wed, 28 Mar 90 16:21:00 CST <UCCXKVB@OSUCC> said: > The best way to get details on this is to get information from your IBM rep (don't ask for a verbal explanation; ask for WRITTEN documentation. There are actually some very good explanations!) >How do you attach an asynch. network to the AIX? A PS/2 running AIX serves as an asynchronous front end terminal server. Think of it as an asynchronous controller. It can actually RUN many of the terminal intensive applications for the AIX/370 such as editors. Limitation: only 16 sessions per PS/2, but you can add up to 30 of them depending on your hardware config. Limitation 2: You have to work through the details of connecting the PS/2 's to the 3090. THis is a Transparent Computing Facility set-up. We intend to use ethernet. Your choice may be different. >We would also like to do AIX/370 from a 3270 terminal. >Is this possible in full screen mode or is only available >in line mode. > Right now, 3270 connection is not possible AT ALL. >The IBM FTN last November discussing AIX/370 and TCF mentioned that >doing vi on the 3090 was very compute intensive since the 370 >architecture does not handle UNIX full screen applications very well. >How much of a resource drain are such UNIX full screen applications? >How much of a resource drain is AIX in general (just to have it up >and running. Other comments are also appreciated.) You do NOT want a 3090 stopping its work to field an interrupt containing a single character. For many users this becomes a VERY EXPENSIVE terminal controller; therefore the use of the PS/2 to provide terminal connection and TCF to do block transfers into the compute engine. In fact, vi can be configured to run on the PS/2 and never on the 370, while the cc runs on the 3090. The file system is a single image between the two. These are short interpretations of the documents. Again, your best bet to answer YOUR SPECIFIC questions is to get the definitive documentation.
DAVID@NERVM.BITNET (David Nessl, Univ. of Fla. -- NERDC) (03/30/90)
On Wed, 28 Mar 90 16:21:00 CST <UCCXKVB@OSUCC> said: >When and if AIX/370 gets announced, Oklahoma State University, >is planning to run it on an IBM 3090-200s. We have a few >questions with AIX/370 with regard to connectivity. >How do you attach an asynch. network to the AIX? >For that matter, how do you connect any asynch term. to AIX. You can't do it thru an FEP -- it must be via TCP/IP telnet, meaning ethernet or (heaven forbid) tokenring. This means that ASCII tubes have to be gatewayed thru one of: the PS/2 setup that IBM recommends, which has additional expense and limits on number of connections; telnet terminal servers, which has the drawback of character-at-a-time I/O to the mainframe; or thru some other generic ASCII host with a telnet service. While we plan to allow limited access via the last two methods for ASCII tubes, we're really planning to promote use of the horde of existing Unix workstations on campus for local editing, with NFS cross-mounts and rsh for execution on the mainframe (fortunately most of our usage will be by researchers who have their own workstations). >We would also like to do AIX/370 from a 3270 terminal. >Is this possible in full screen mode or is only available >in line mode. You can get linemode-only access after logging on to VM or TSO and telnetting to the AIX/370 system. Unix full-screen capability is not possible from a 3270 because 3270's can't do character-at-a-time keyboard input (it's not due to host software limitations). >The IBM FTN last November discussing AIX/370 and TCF mentioned that >doing vi on the 3090 was very compute intensive since the 370 >architecture does not handle UNIX full screen applications very well. >How much of a resource drain are such UNIX full screen applications? >How much of a resource drain is AIX in general (just to have it up >and running. Other comments are also appreciated.) We've heard comments from other universities that it really isn't that bad. We're taking a wait-and-see attitude: if it really is that bad, we can purchase or reallocate PS/2's fast enough. Plus our 8232 is on it's own chpid, so the I/O will only affect TCP/IP, and I'm betting that the 8232 will bottleneck before the 3090 does. We'll see. -david David Nessl, systems programmer & postmaster BITNET: david@nervm |Northeast Regional Data Center Internet: david@nervm.nerdc.ufl.edu |233 SSRB, University of Florida Voice: 904 392 4601 |Gainesville, FL 32611 USA
MAB@CORNELLC.BITNET (Mark Bodenstein) (03/30/90)
On Thu, 29 Mar 90 09:08:24 EST Alton Brantley said: > ... > You do NOT want a 3090 stopping its work to field an interrupt containing >a single character. It's not (quite) as bad as first appears. Packets for the 3090 are buffered at the 8232 for a short while, to see if they can be aggregated with other packets. Under load, this means that there will be only 1 I/O interrupt on the 3090 to receive many packets from many different sources. This helps quite a bit. Mark Bodenstein (mab@cornellc; 607-255-3747) Cornell University