andyd@pogo.WV.TEK.COM (Andy Davidson) (01/10/91)
We are in the process of planning the installation of some Farralon PhoneNET
StarControllers to relieve some overloaded net problems. For each of the 12
ports on a single controller, there are 4 "branches" each of which can have
up to 750 feet of cabling. Farralon technical support says that they
recommend and support only one device (Mac, printer, whatever) per branch.
Given that each port is electrically isolated from the others, this seems
extremely conservative.
Hence my question: Have any of you installed this controller and what do you
find are the *practical* limits on devices per branch?
thanks,
andy
--
Andy Davidson 503-685-3033 Manager, Tektronix/GPID QA & Tools
Internet: andyd@pogo.WV.tek.com uucp: ...!tektronix!pogo!andyd
Trust, but verify!roy@phri.nyu.edu (Roy Smith) (01/11/91)
andyd@pogo.WV.TEK.COM (Andy Davidson) writes: > Farralon technical support says that they recommend and support only one > device (Mac, printer, whatever) per branch. Given that each port is > electrically isolated from the others, this seems extremely conservative. When building networks, being conservative may not be a bad idea. At any rate, I usually allow several devices per branch, as long as all the devices are right next to each other in the same room and can be connected with just a short piece of PhoneNet modular cable, and are in a strictly linear chain with a single terminator at the end. I consider branches beyond the StarController's punchdown block to be strictly verbotten. It seems to work. Keep in mind, however, that if you start violating Farallon's recommendations, even though you may very well get away with it, you are living on borrowed time. The most we have on a branch right now is a FastPath, a LaserWriter, and three macs. One of these days, the FastPath is going to get taken off that branch and given an entire port all to itself, just to make it easier to troubleshoot network problems (i.e. I can turn off that one port without effecting anything else). -- Roy Smith, Public Health Research Institute 455 First Avenue, New York, NY 10016 roy@alanine.phri.nyu.edu -OR- {att,cmcl2,rutgers,hombre}!phri!roy "Arcane? Did you say arcane? It wouldn't be Unix if it wasn't arcane!"
awessels@ccwf.cc.utexas.edu (Allen Wessels) (01/11/91)
In article <10299@pogo.WV.TEK.COM> andyd@pogo.WV.TEK.COM (Andy Davidson) writes: >We are in the process of planning the installation of some Farralon PhoneNET >StarControllers to relieve some overloaded net problems. For each of the 12 >ports on a single controller, there are 4 "branches" each of which can have >up to 750 feet of cabling. Farralon technical support says that they >recommend and support only one device (Mac, printer, whatever) per branch. >Given that each port is electrically isolated from the others, this seems >extremely conservative. I'm sure that the fact that they'll make serious $ by selling one SC per 12 devices doesn't enter the picture at all. (Actually, they want to sell 1 per 11, with the 12th port used to connect to the next SC. "How conveeeenient.") Seriously, IMO, you can hang as many devices off the SC as you want, keeping in mind bandwidth limitations. In what way do you plan on using the SC to relieve net "overload"? The SC is basically just a big, fancy repeater. If you are having traffic problems you need to look into a bridge. >Hence my question: Have any of you installed this controller and what do you >find are the *practical* limits on devices per branch? I have 1 SC on a net with 35-odd Macs, 1 PC, 4 LaserWriters, and 3 AppleTalked ImageWriter IIs. There is one AppleShare server on a Mac Plus that two Pluses using 4th Dimension are grabbing a 15 meg datebase in multiuser mode from. All this works pretty well. In the 2-3 years I've had this setup running, I've had to use the StarCommand softare *twice* to shut down a leg. I'd guess 6-8 devices per port would be a fair rule of thumb upper limit. This is an average. It wouldn't bother me much to hang a dozen off a leg if I needed to for some reason.
wards@iies.ecn.purdue.edu (Sherman Ward) (01/11/91)
In article <1991Jan10.193750.25098@phri.nyu.edu> roy@phri.nyu.edu (Roy Smith) writes: >andyd@pogo.WV.TEK.COM (Andy Davidson) writes: >> Farralon technical support says that they recommend and support only one >> device (Mac, printer, whatever) per branch. Given that each port is >> electrically isolated from the others, this seems extremely conservative. > > When building networks, being conservative may not be a bad idea. >At any rate, I usually allow several devices per branch, as long as all the >devices are right next to each other in the same room and can be connected >with just a short piece of PhoneNet modular cable, and are in a strictly >linear chain with a single terminator at the end. I consider branches >beyond the StarController's punchdown block to be strictly verbotten. It >seems to work. Keep in mind, however, that if you start violating >Farallon's recommendations, even though you may very well get away with it, >you are living on borrowed time. The most we have on a branch right now is >a FastPath, a LaserWriter, and three macs. One of these days, the FastPath >is going to get taken off that branch and given an entire port all to >itself, just to make it easier to troubleshoot network problems (i.e. I can >turn off that one port without effecting anything else). >-- >Roy Smith, Public Health Research Institute >455 First Avenue, New York, NY 10016 >roy@alanine.phri.nyu.edu -OR- {att,cmcl2,rutgers,hombre}!phri!roy >"Arcane? Did you say arcane? It wouldn't be Unix if it wasn't arcane!" Andy & Roy, I missed the original article, so ..... I will ad my 2 cents about the StarController Branches. We installed LocalTalk in 1987 (using twisted pair coax and Apple's connector kits). Our department acquired a StarController after our LocalTalk net became too long & un-reliable. (We tried 2 Tops Repeaters first, but it was a short term fix :-( Our decision to buy a StarController was aided by the existence of old "dumb" terminal port wiring in the walls. We placed the Controller next to the punch block and used the old ports to create 8 branches. Some of the branches have only 1 "unit" on the branch. All of the old cabling was used in conjunction with the new arrangement. We had to buy 12 converters (going from RJ11 to Apple's 3 pin LocalTalk) and the rest was 'plug & play'. Our most populated branch has 4 Macs, LaserWriterII, and a Kennetics (Shive) FastPath. Another branch has 5 Macs and a LaserWriterII. Baring janitors mops, clumsy feet, & sheer ignorance which disconnect the LocalTalk pieces, the StarController has worked flawlessly :-) (..YEA RAH something that has worked EXACTLY as advertised ;-) I agree with Roy in that the "net access box" (FastPath, GatorBox, etc..) sould be the only device on one branch because of the "importance" of it's services. (Our arrangement didn't allow this :-( I also understand the 'conservative' approach, but "bang for the buck" doesnt allow you to be TOO conservative. Just my 2 cents.. Sherman -- +SNAIL: Sherman Ward / Freshman Engineering / Purdue University / + + Rm 221 Engineering Admin. Bldg. / West Lafayette, IN 47907 + +-------------------------+----------------+------------------------+ + wards@pc.ecn.purdue.edu + (317) 494-3976 + I believe in Murphy! +
roy@phri.nyu.edu (Roy Smith) (01/11/91)
awessels@ccwf.cc.utexas.edu (Allen Wessels) writes: > I'm sure that the fact that they'll make serious $ by selling one SC per 12 > devices doesn't enter the picture at all. (Actually, they want to sell 1 per > 11, with the 12th port used to connect to the next SC. "How conveeeenient.") I am pretty sure the limitation is one device per *branch*, and you are allowed 4 branches per port (one on each of the 4 pairs of punchdown terminals). That makes it 48 (or 44) devices per SC. By the time you have that many devices on one LocalTalk segment, you will probably want to start looking into traffic-filtering bridges to segregate your network into smaller pieces anyway, so I don't consider the device-count limitation to really be that severe. -- Roy Smith, Public Health Research Institute 455 First Avenue, New York, NY 10016 roy@alanine.phri.nyu.edu -OR- {att,cmcl2,rutgers,hombre}!phri!roy "Arcane? Did you say arcane? It wouldn't be Unix if it wasn't arcane!"
awessels@ccwf.cc.utexas.edu (Allen Wessels) (01/12/91)
In article <1991Jan11.135537.22335@phri.nyu.edu> roy@phri.nyu.edu (Roy Smith) writes: > I am pretty sure the limitation is one device per *branch*, and you >are allowed 4 branches per port (one on each of the 4 pairs of punchdown >terminals). That makes it 48 (or 44) devices per SC. By the time you have I haven't looked at the documentation for the SC in a long time, but if I remember right, only the ports are individually controllable, not the branches, so for ideal control on the network...(I decided to look it up) ---------------------- page 33 of the SC user's guide Recommended Cabling Configuration For easy network management and troubleshooting, we recommend that you connect one branch to each StarController Port and one device to each branch. Using this configuration, you can connect up to 12 branches and 12 devices (one device per branch) to a single StarController. This configuration lets you quickly identify a network device that is malfunctioning and prevent the malfunctioning device from affecting other devices. ---------------------- >that many devices on one LocalTalk segment, you will probably want to start >looking into traffic-filtering bridges to segregate your network into >smaller pieces anyway, so I don't consider the device-count limitation to >really be that severe. There isn't any advantage to hooking up one device per port branch on the SC, so I use the 3000 ft limit per port and hook up as many devices as I need to. My network is fairly low traffic, and I figure I could go to well over 44 devices on our SC. My advice is to ignore Farallon's recommendations and, following their limits on wiring distances and branching, design your network after having mapped out both the physical layout and the likely data flow over the network. I've seen situations in which 25 machines should be considered a maximum and others where 60 was tolerable.
derosa@motcid.UUCP (John DeRosa) (01/12/91)
andyd@pogo.WV.TEK.COM (Andy Davidson) writes: > ...............For each of the 12 >ports on a single controller, there are 4 "branches" each of which can have >up to 750 feet of cabling. Farralon technical support says that they >recommend and support only one device (Mac, printer, whatever) per branch. We have 20 or so star controllers in our network. Each of the branches on each star controller has more than one device, in fact four "runs" to an office per branch is standard. From time to time, one of those runs magically "grows" another device and all is still well. Now, all this is based on experience rather than testing. I assume that as more devices are added to a zone, the load increases and throughtput decreases. Another of our divisions did some testing and stated that 15 devices is the magic number, more than that and your throughput decreases. -- = John DeRosa, Motorola, Inc, Cellular Infrastructure Group = = e-mail: ...uunet!motcid!derosaj, motcid!derosaj@uunet.uu.net = = Applelink: N1111 = =I do not hold by employer responsible for any information in this message =
woody@ucscb.UCSC.EDU (Bill Woodcock) (01/14/91)
> We are in the process of planning the installation of some > Farallon PhoneNET StarControllers to relieve some overloaded > net problems. For each of the 12 ports on a single > controller, there are 4 "branches" each of which can have up > to 750 feet of cabling. Farralon technical support says > that they recommend and support only one device (Mac, > printer, whatever) per branch. Given that each port is > electrically isolated from the others, this seems extremely > conservative. > > Hence my question: Have any of you installed this > controller and what do you find are the *practical* limits > on devices per branch? To summarize the _Farallon Party Line_... StarControllers have twelve ports, each of which can best support one device at the end of <3000' of 24ga. solid core twisted pair IC phone cable. The _maximum_ rated spec for a StarController is 48 devices. Dividing 48 by 12 yields 4 devices per port. Dividing 3000' by 4 devices yields 750'. Thus the maximum allowable within spec is four devices per port, each at the end of <750' of cable. As you've correctly surmised, the electrical isolation of each port makes each one a separate physical (if not logical) network. It's relatively easy to address your question from this standpoint. What Farallon is saying is exactly what they'd tell you if you were building twelve separate extrememly small networks without a StarController. You can use up to a grand total of 3000' of cable in your network, and if you're going to make a passive star topology with it, you'd better not try to use more than four branches, since signal strength on a passive star is divided by the number of branches. The little networks you're planning on attaching to each of the ports of your StarController can be treated as individual passive stars (evil, evil, evil...) and should follow the same rules. In addittion, there's the problem of traffic. 48 Macs and gateways and LaserWriters generate a heck of a lot of traffic on a baseband network, and if you actually did put that many nodes in one zone, odds are you wouldn't be a happy person for the experience. Farallon has always recommended about 45 nodes per zone, traffic-wise, and Apple has always said 32. The figure is going to actually depend on the amount of traffic your particular users generate. Kurt Vandersluis published a formula for roughing it out in MacAbuser around last August sometime. In answer to your final question, I'd say from my own experience that you would do well to stick to either one or two nodes per port, but that you might be able to get up to 4000' of cable, if you're careful laying it, and avoid intermediary punchdown blocks as much as possible. It's also possible to use 22ga. cable, to increase the possible length of your cable segments. -Bill Woodcock BMUG NetAdmin _______________________________________________________________________________ 0000 : 0600 0800 7700 FE00 FF 0 FF8 7F 0 3 00 48 bill.woodcock.iv 0010 : CC00 7C00 0C00 1000 2800 440 8200 40 0020 : C000 4000 8000 C000 C 00 800 80 48 0 9 woody@ucscb.ucsc.edu 0030 : FF00 F000 7F80 CC00 CC 0 7 80 7 80 CC 6 0040 : CC00 7F80 9800 7800 CC00 CC0 CC 0 C 00 2355.virginia.st 0050 : CC00 CC00 FC00 CC00 CC 0 2000 10 0 7 00 C 0 0060 : CC00 CC00 CCC0 4800 2 00 1 00 2 00 48 0 9 berkeley.california 0070 : 4800 2400 1200 1000 2800 4 00 8 00 E0 0080 : 3000 6000 F000 6000 000 60 0 600 C0 0 01C 94709.1315 _______________________________________________________________________________