crum@fcom.cc.utah.edu (Gary Crum) (03/21/91)
Would anyone care to share stories (problems or success reports) about using AIX 3.1 BNU (UUCP), especially with 9600bps or faster modems? I have had a few problems with a cross-country BNU link between a couple RISC System/6000 computers using Courier HST Dual Standard modems. Throughput is around 1500bps, and connections are frequently broken. That's with a recently installed 3.1.2 release of AIX -- I'm not sure what week-format version number I'm using. Also, has anyone used SLIP for file transfers, especially using the dialout feature of slattach? How about BNU (UUCP) on top of SLIP connection? Can BNU be configured to establish SLIP connections before trying to do TCP UUCP transfers? Just wondering if there's much experience out there on the net with these issues -- before I try to talk to Austin about it. Gary
robin@pensoft.UUCP (Robin Wilson) (03/27/91)
In article <CRUM.91Mar20185927@fcom.cc.utah.edu> crum@fcom.cc.utah.edu (Gary Crum) writes: >Would anyone care to share stories (problems or success reports) about >using AIX 3.1 BNU (UUCP), especially with 9600bps or faster modems? I >have had a few problems with a cross-country BNU link between a couple >RISC System/6000 computers using Courier HST Dual Standard modems. >Throughput is around 1500bps, and connections are frequently broken. >That's with a recently installed 3.1.2 release of AIX -- I'm not sure >what week-format version number I'm using. Get the 3003 update and apply it to your system. There were several tty bugs that are now fixed, and several uucico/uucp/uux bugs that were fixed in the newer updates. I have previously posted articles about setting up modems on the RS, so you can either get the old article from an archive or mail me directly and I will send you a description. >Also, has anyone used SLIP for file transfers, especially using the >dialout feature of slattach? How about BNU (UUCP) on top of SLIP I have used it, and it is broken also until 3004. The major problem was with multiple slip connections, so if you only have 1, it should work OK. BTW, I had to do it manually to get it to work. SMIT always gave me too many other problems. Mail me if you need help with this. >connection? Can BNU be configured to establish SLIP connections before >trying to do TCP UUCP transfers? Why would you want to do this? Why don't you just use rcp or something? +-----------------------------------------------------------------------------+ |The views expressed herein, are the sole responsibility of the typist at hand| +-----------------------------------------------------------------------------+ |UUCP: pensoft!robin | |USNail: 701 Canyon Bend Dr. | | Pflugerville, TX 78660 | | Home: (512)251-6889 Work: (512)343-1111 | +-----------------------------------------------------------------------------+
jkimble@unislc.uucp (Jim Kimble) (03/28/91)
In article <...> crum@fcom.cc.utah.edu (Gary Crum) writes: >Would anyone care to share stories (problems or success reports) about >using AIX 3.1 BNU (UUCP), especially with 9600bps or faster modems? I >have had a few problems with a cross-country BNU link between a couple >RISC System/6000 computers using Courier HST Dual Standard modems. >Throughput is around 1500bps, and connections are frequently broken. I'm the dork who did the port of BNU UUCP to the RS/6000 while contracting at IBM Austin. For grins and giggles, I connected a couple TrailBlazers (T1500s, I believe) and did some throughput measurements. My figures between two RS/6000's were a *lot* better than 1500bps but didn't compare to the figures my friends/neighbors were seeing between NCR Towers, DELL boxes, and TI BS1500s. The real killer is connecting to an IBM RT because VRM had a nasty problem with high-speed UUCP connections overflowing the i/o buffers -- consequently VRM handled the overflow by sending either a SIGHUP or SIGINT to uucico which caused the connection to die for mysterious reasons... The "fix" for that APAR was to slow down uucico rather than speed up VRM. No shit. What's your system configuration? What else is running on the system (number of users, applications, etc)? DISCLAIMER: This article does not represent the thoughts of *any* company or even those of a sane mind. Use them at your risk. -- --Jim Kimble, jkimble@bally.bally.com Yet Another UNIX Insultant sun!unislc!jkimble "ALPO is 99 cents a can. That's over SEVEN dog dollars!"
crum@fcom.cc.utah.edu (Gary Crum) (03/29/91)
In article <3354@pensoft.UUCP> robin@pensoft.UUCP (Robin Wilson) writes: >connection? Can BNU be configured to establish SLIP connections before >trying to do TCP UUCP transfers? Why would you want to do this? Why don't you just use rcp or something? It is often appropriate to queue up transfer requests and do them in bulk, automatically, at scheduled times. UUCP does such queueing and offers control over connection scheduling (e.g. the UUCP Systems file specifies what times are appropriate to establish connections). UUCP also allows polling to be setup. Most of the IP/TCP user-level programs, like rcp, were designed with dedicated links in mind, not dialup SLIP links. UUCP was designed with dialup links in mind, not unlike dialup SLIP links. Gary
mdapoz@hybrid.UUCP (Mark Dapoz) (03/30/91)
In article <1991Mar28.110956.10576@unislc.uucp> sun!jkimble@unislc (Jim Kimble) writes: >I'm the dork who did the port of BNU UUCP to the RS/6000 while contracting >at IBM Austin. For grins and giggles, I connected a couple TrailBlazers >(T1500s, I believe) and did some throughput measurements. My figures >between two RS/6000's were a *lot* better than 1500bps but didn't compare >to the figures my friends/neighbors were seeing between NCR Towers, DELL >boxes, and TI BS1500s. The most we've been able to get using either 'blazers or HST's is around 4800 baud effective throughput, which is pretty pittiful. This was between two 6000's at level 3003. >The real killer is connecting to an IBM RT because VRM had a nasty problem >with high-speed UUCP connections overflowing the i/o buffers -- consequently >VRM handled the overflow by sending either a SIGHUP or SIGINT to uucico >which caused the connection to die for mysterious reasons... The "fix" for >that APAR was to slow down uucico rather than speed up VRM. No shit. So why in the world would you leave a nap() call between each packet when you ported uucico to the 6000! This effectively kills any chance of doing high speed communication with uucp. We've got an apar open on this so hopefully it'll get fixed soon. Sheesh, any you even knew about it. -- Mark Dapoz home: mdapoz%hybrid@cs.toronto.edu work: md@toronto.ibm.com or mdapoz@torvm3.vnet.ibm.com Finger and toes, finger and toes, forty things we share, forty one if you include the fact that we don't care. - The Tragically Hip
karish@mindcraft.com (Chuck Karish) (04/02/91)
In article <1991Mar30.032817.7885@hybrid.UUCP> mdapoz@hybrid.UUCP (Mark Dapoz) quotes <1991Mar28.110956.10576@unislc.uucp> sun!jkimble@unislc (Jim Kimble): >>The real killer is connecting to an IBM RT because VRM had a nasty problem >>with high-speed UUCP connections overflowing the i/o buffers -- consequently >>VRM handled the overflow by sending either a SIGHUP or SIGINT to uucico >>which caused the connection to die for mysterious reasons... The "fix" for >>that APAR was to slow down uucico rather than speed up VRM. No shit. This is the first I've heard about a "fix" for this one, and I've been bugging the support people about it for two years. If I have the 6150 talk to my TrailBlazer at 19200 baud, it loses characters. The machines has a buffered 4-port adapter (16550 UART chips) and it works at 9600 Baud max, which gives me 700-800 cps on UUCP transfers. The UUCP symptom is that uucico times out and kills itself with an assert(), telling me that it's at line 365 in program pk1.c Chuck Karish karish@mindcraft.com Mindcraft, Inc. (415) 323-9000
markw@airgun.wg.waii.com (Mark Whetzel) (04/02/91)
In article <670532360.4967@mindcraft.com>, karish@mindcraft.com (Chuck Karish) writes: > In article <1991Mar30.032817.7885@hybrid.UUCP> mdapoz@hybrid.UUCP (Mark Dapoz) > quotes <1991Mar28.110956.10576@unislc.uucp> sun!jkimble@unislc (Jim Kimble): [ discussions of VRM and ect, deleted ] > > This is the first I've heard about a "fix" for this one, and I've been > bugging the support people about it for two years. If I have the 6150 > talk to my TrailBlazer at 19200 baud, it loses characters. The > machines has a buffered 4-port adapter (16550 UART chips) and it works > at 9600 Baud max, which gives me 700-800 cps on UUCP transfers. > > The UUCP symptom is that uucico times out and kills itself with an > assert(), telling me that it's at line 365 in program pk1.c > I currently use a T1500 trailblazer for my uucp connections to uunet all the time without any loss of characters. I am not sure of performance, but I have never measured thruput (or how to find out :-) I am running on an IBM RT 135 running AIX 2.2.1 at maint level 2706+1773 using the internal serial port (S2) for 19200 baud connection just fine. I have it setup to use DMA channel zero. It works like a charm! Here is excerpt from my devices setup: Current Possible Name Description Choice Choices biopa First I/O Port Address 8001 8001 cn DMA Channel Number 0 0 <<<NOTE! ae Automatic Enable delay true,false,share,delay bpc Bits Per Character 8 5,6,7,8 pt Parity Type none even,odd,none rts Receive/Transmit Speed 19200 pro Protocol dc dtr,cdstl,dc ixp Include Xon/Xoff Protocol false true,false dvam Device Attachment Method 1 0 = local, 1 = remote (modem) noi Num. of Interupt Levels Used 2 2 pn Port Number on Adapter 0 0 dmas DMA Support true true,false <<<NOTE! sdmac Shared DMA Channel false true,false ei1 Enable 1st Interrupt Level true true,false si1 Share 1st Interrupt Level true true,false ic1 Service Class of 1st Interpt 0 0,1,2,3 il1 Intrpt Level # of 1st Intrpt 2 2 <<< NOTE! ei2 Enable 2nd Interrupt Level true true,false si2 Share 2nd Interrupt Level true true,false ic2 Service Class of 2nd Interpt 0 0,1,2,3 il2 Intrpt Level # of 2nd Intrpt 13 13 I have heard of performance problems with both the 4-port and 8-port adapters from IBM. If you look in the "User setup guide and options installation" book, on page 1-21 you will see that the internal ports use DMA channels AND interrupts for I/O (it uses IRQ 2 and 13, which is reserved for just this purpose!) whereas the other ASYNC cards use only interrupts! No wonder they overflow especially on a hi-speed link and with any users on to delay service of the cards! Later, markw -- Mark Whetzel My comments are my own, not my company's. Western Geophysical - A division of Western Atlas International, A Litton/Dresser Company DOMAIN addr: markw@airgun.wg.waii.com UUNET address: uunet!airgun!markw