eggers@ndmath.UUCP (Mark Eggers) (03/28/89)
I have the following configuration set up, which seems to be 'legal' according to DEC's Network and Communications Buyer's Guide. +----------------+ | DELNI | +----------------+ | | | | | | | | | | | | | | | +---p4200 router | | | | | | | | | | | | | +-----ACS 9310-----IBM 3033 | | | | | | | | | | | +-------Annex terminal server | | | | | | | | | +---------Cayman GatorBox | | | | | | | +-----------Mac II running A/UX (only, no Mac OS) | | | | | +-------------microVAX II running Ultrix 2.2 | | | +---------------Compaq 286 running Excelan's LanAlyzer | +-----------------DESPR (one port version of a DEMPR) | | +------Convex C120 | +------Convex C120 | +------Convex C220 The problem: As long as I telnet, rlogin, ftp, etc. between the three Convexes on the thinwire Ethernet, performance is what one would expect (no noticable delays in telnet IO, 50 - 100 K file transfer rates). As soon as one of the Convexes attempts to go through the DESPR, performance becomes 'terrible' (delays of one to two seconds at odd times in telnet IO, ftp of 3 - 10 K). Some other facts 1. Performance on the 'DELNI' side is reasonable (no noticable delays in telnet IO, ftp of 35 - 100 K). 2. Replacing a Convex with the Compaq does not change the results. Compaq performance to devices on the thinwire Ethernet are fine, through the DESPR are terrible. 3. A Convex directly on the DELNI performs well. Notes on configuration 1. We run both DECnet and TCP/IP. The microVAX II is the DECnet - TCP/IP gateway for the campus, and there are DNA routing packets zipping about on the local net. 2. The p4200 is set to broadcast RIP on this subnet, and the microVAX II runs /etc/routed. Both are probably unnecessary. 3. All devices understand subnets and are configured for it. 4. No other device (except the microVAX II) is running either /etc/routed or rwhod. 5. All devices on the thinwire Ethernet are connected via DESTAs (thinwire transceivers) and 50' of transceiver cable. The thinwire Ethernet is properly terminated. Any thoughts on this problem would be greatly appreciated. Right now, we have removed the terminal server, DESPR, and Compaq in order to make room for the three Convexes. Needless to say, I would like all of these devices (plus 2 NeXTs) on the net. /mde/
eriks@yunexus.yorku.ca (Eriks Rugelis) (03/28/89)
>I have the following configuration set up, which seems to be 'legal' >according to DEC's Network and Communications Buyer's Guide. 'seems to be legal'?... I haven't seen a recent copy of the N&CBG but I would be surprised if DEC allowed you to plug a DESPR into a DELNI why? well because the DESPR is a repeater and a DEC supplied DELNI operating in 'local' mode generates heartbeat and repeaters cannot tolerate heartbeat; a repeater operating in the presence of heartbeat will see excessive collisions and will have the poor performance behaviour that you describe solution? replace the DELNI with a similar box not made by DEC that allows you to select/deselect heartbeat (i.e. I use Cabletron's MT-800's)... I have installed and running many instances of repeaters plugged into multi-port transceivers so it is demonstrably workable Eriks --- Voice: Eriks Rugelis E-Mail: eriks@libra.yorku.ca Ma Bell: 416/736-5257 x.2688 UUCP: yunexus!eriks Last resort: eriks@yulibra.NETNORTH Snail: York University, 4700 Keele St., North York, Ontario, Canada. M3J 1P3 Disclaimer: I have no interest in Cabletron outside of being a satisfied customer. -- Voice: Eriks Rugelis E-Mail: eriks@libra.yorku.ca Ma Bell: 416/736-5257 x.2688 UUCP: yunexus!eriks Last resort: eriks@yulibra.NETNORTH Snail: York University, 4700 Keele St., North York, Ontario, Canada. M3J 1P3
rpw3@amdcad.AMD.COM (Rob Warnock) (03/28/89)
In article <ERIKS.89Mar27222745@yunexus.yorku.ca> (Eriks Rugelis) writes: +--------------- | >I have the following configuration set up, which seems to be 'legal' | >according to DEC's Network and Communications Buyer's Guide. | ...I would be surprised if DEC allowed you to plug a DESPR into a DELNI | why? well because the DESPR is a repeater and a DEC supplied DELNI operating | in 'local' mode generates heartbeat and repeaters cannot tolerate heartbeat; | a repeater operating in the presence of heartbeat will see excessive | collisions and will have the poor performance behaviour that you describe +--------------- Almost. What a repeater cannot tolerate is a multi-port which does heartbeat *incorrectly*, such as a DELNI. Doing heartbeat "correctly" here means that you only give a heartbeat [little burst of "collision" immediately following a transmitted packet] to the port which transmitted. I believe that DELNI's give back a heartbeat [heartburn?] to *all* ports when *any* port transmits. This is simply carelessness on the part of the designer(s). It takes all of $0.50 worth of logic per port (a non-critical one-shot) to do it right. +--------------- | solution? replace the DELNI with a similar box not made by DEC that allows | you to select/deselect heartbeat (i.e. I use Cabletron's MT-800's)... +--------------- ...or a box which does it correctly [by the above definition]. I'm not sure if the Cabletron does it right, either, but at least they let you disable heartbeat per port (which is enough). Rob Warnock Systems Architecture Consultant UUCP: {amdcad,fortune,sun}!redwood!rpw3 DDD: (415)572-2607 USPS: 627 26th Ave, San Mateo, CA 94403
abstine@sun.soe.clarkson.edu (Arthur Stine) (03/28/89)
From article <ERIKS.89Mar27222745@yunexus.yorku.ca>, by eriks@yunexus.yorku.ca (Eriks Rugelis): >>I have the following configuration set up, which seems to be 'legal' >>according to DEC's Network and Communications Buyer's Guide. > 'seems to be legal'?... I haven't seen a recent copy of the N&CBG but > I would be surprised if DEC allowed you to plug a DESPR into a DELNI > > why? well because the DESPR is a repeater and a DEC supplied DELNI operating > in 'local' mode generates heartbeat and repeaters cannot tolerate heartbeat; > a repeater operating in the presence of heartbeat will see excessive collisions > and will have the poor performance behaviour that you describe > > solution? replace the DELNI with a similar box not made by DEC that allows > you to select/deselect heartbeat (i.e. I use Cabletron's MT-800's)... > I have installed and running many instances of repeaters plugged into > multi-port transceivers so it is demonstrably workable > > Eriks Almost...according to the 'book', you can attach up to eight DEMPR/DESPR to a DELNI provided: 1) the standard ethernet segment the DELNI is attached to is < 300 meters and it is attached using a H4000-BA/H4005 (standard transceiver w/o heartbeat) 2) if it is standalone, put the DELNI in global mode and put a loopback connector on the network port. This will eliminate the heartbeat signal, as the DELNI will take heartbeat from whatever it is attached to when it is in global mode. Note that the Cabletron MT800 allows you to turn on/off heartbeat in a much easier way (a switch), but it still does not allow you to do it on a port by port basis. art stine sr network engineer clarkson u
ted@blia.BLI.COM (Ted Marshall) (03/29/89)
Another way to correctly connect the DESPR to the DELNI would be to put the DELNI into "global" mode (such that it acts as an 1-to-8 fan-out instead of an Ethernet-in-a-box). Then connect the DELNI global port to a transceiver, connected to a 2.5 meter slice of thick cable, connected to a no-heartbeat transceiver, connected to the DESPR. This way, you don't have to throw away your DELNI and buy something else. -- Ted Marshall ...!ucbvax!mtxinu!blia!ted <or> ted@blia.bli.com Britton Lee, Inc., 14600 Winchester Blvd, Los Gatos, Ca 95030 (408)378-7000 The opinions expressed above are those of the poster and not his employer.
koblas@mips.COM (David Koblas) (03/29/89)
In article <7117@blia.BLI.COM> ted@blia.BLI.COM (Ted Marshall) writes: > >Another way to correctly connect the DESPR to the DELNI would be to put the >DELNI into "global" mode (such that it acts as an 1-to-8 fan-out instead of >an Ethernet-in-a-box). Then connect the DELNI global port to a transceiver, >connected to a 2.5 meter slice of thick cable, connected to a no-heartbeat >transceiver, connected to the DESPR. > >This way, you don't have to throw away your DELNI and buy something else. A much simpler way to go about the same thing is to connect a thinnet tranciever (like a DESPR but not a repeater) to the Tranciever port on the DELNI a and then just run thinnet. Then all you have to scrap is the DESPR. -- name : David Koblas uucp : {ames,decwrl}!mips!koblas place: MIPS Computers Systems domain: koblas@mips.com quota: "It was never ment to be a game, NEVER." -- Rollerball
dd@ariel.unm.edu (03/30/89)
In article <1349@ndmath.UUCP> eggers@ndmath.UUCP (Mark Eggers) writes: >I have the following configuration set up, which seems to be 'legal' >according to DEC's Network and Communications Buyer's Guide. [...] >The problem: > > As long as I telnet, rlogin, ftp, etc. between the three >Convexes on the thinwire Ethernet, performance is what one would >expect (no noticable delays in telnet IO, 50 - 100 K file transfer rates). > > As soon as one of the Convexes attempts to go through the >DESPR, performance becomes 'terrible' (delays of one to two seconds >at odd times in telnet IO, ftp of 3 - 10 K). [...] >Any thoughts on this problem would be greatly appreciated. Right >now, we have removed the terminal server, DESPR, and Compaq in >order to make room for the three Convexes. Needless to say, I would >like all of these devices (plus 2 NeXTs) on the net. We have had performance problems like this when we have mixed E'net V.1 and no-heartbeat V.2 on a DELNI with heartbeat V.2 and IEEE 802.3 accidentally. Have you kept careful track of this parameter? If you still can't figure it out, and if you have the $$$ to play with it, try substituting Cabletron eqpt in for the DEC eqpt. They have a plethora of diagnostic LEDs they call "LANVIEW" which really help to figure out what's going on. Don Doerner (dd@ariel.unm.edu) University of New Mexico CIRT 2701 Campus Blvd. NE. Albuquerque, NM, 87131 (505) 277-8036 Don Doerner (dd@ariel.unm.edu) University of New Mexico CIRT 2701 Campus Blvd. NE.
jeff@nsipo.arc.nasa.gov (Jeff Burgan) (04/01/89)
In article <1349@ndmath.UUCP> eggers@ndmath.UUCP (Mark Eggers) writes: > As soon as one of the Convexes attempts to go through the >DESPR, performance becomes 'terrible' (delays of one to two seconds >at odd times in telnet IO, ftp of 3 - 10 K). > Since a lot of other's have already explained the repeater and heartbeat problem, all I will add is they they are correct. A trick that I learned about not too long ago that no one has mentioned is to put the DELNI in remote mode and put an ethernet test plug in the remote port. The test plug (probably not its official name) comes with a DESPR or DEMPR. Hope this helps, Jeff