jsa00564@uxa.cso.uiuc.edu (02/18/89)
In article <98@cs.columbia.edu> benderly@cs.columbia.edu (Dan Benderly) writes: >Does anyone have any experiences with running Ethernet over unshielded >twisted pair ? I know that Lattisnet has such a product, and I was wondering >what people's impressions were. I am a University student who's been lucky enough to do systems integration for a medium sized LAN installation. We installed Lattisnet last August, and have had good luck with it. We did the installation a little differently than most, as we wanted to install 50+ PC's using existing wire and jacks (we just had a new PBX installed last April). Since the jacks installed were dual RJ-11 (2 pair), but Lattisnet uses 4-pair RJ-45 jacks (2 pairs are not used). So, we had some custom RJ11 to RJ45 cables made at Black Box to make the connection. I'd be more than happy to pass on the specs, as it made the installation (from the phone-trained installer) look just like another phone extension (i.e. it keeps the standard wire color code throughout the system). It doesn't sound like a big deal, but there was a lot of mistakes made before we got it right..... Anyway, the system works well, but the installation is finicky. The wire MUST be solid-core copper from end to end, with VERY solid punches at any punch blocks. Also, the cable-length limit of 360 feet is quite strict. (We use Lattisnet "Link Extenders" to go up to 720 feet). Initial installation is the only tough part, though. Day to day, the nodes stay up, with very little problem (at least, problems related to wiring and such...). We used Micom-InterLAN's NI-5210-UTP cards, but others are available. We've had good luck with them, only one defective unit out of 54 units. Please feel free to call me or email any questions.... ^^^^^-preferred ------------------------------------------------------------------------------ Jeff Autor jsa00564@uxa.cso.uiuc.edu "Help me! I'm an engineering student!" University of Illinois Also affiliated with Valmont Electric, Danville, IL (217) 446-4600
pnessutt@nis.mn.org (Robert A. Monio) (02/20/89)
In article <1689@umbc3.UMBC.EDU> digennar@umbc3.umbc.edu.UMBC.EDU (Mr. Jerry DiGennaro) writes: >The firm I work for has two and ahalf buildings running LattisNet at >this time. We have over 15 Novell file servers, 1 VAX 8810, a Wang >7310, and 250+ PCs running on it. We plan to put several Tandems, >several Perkin Elmers, an IBM 4341, and anything else we can think of on >this Ethernet. Could someone send me the address and phone number of the company that vendors LattisNet? Our company is presently looking for some methodology that we could utilize to connect our PC's, Unix Boxes, and 4381 together. We are presently using EXOS 201's to connect our Unix machines, but we are now looking at expanding our ethernet to include PC's, scanners, and the 4381. I'm presently collecting information on companies that vendor Network software and I would be happy if someone could send along the address and phone number to me. On a further note: Has anyone had any luck/success with the SMB File Server from SYNTAX, Inc.? Apparently they have a NFS-Compatible File/Server that will allow us to network our Tower Unix boxes with our PC's over the Ethernet. Why not use NFS instead, you say? Well, we're forced to go this route because NCR hasn't bothered to keep up with the rest of the world in regards to it's Unix Releases. Even though I'm a CVT site for OS:03.00.00, I still regard the timeframes I have for implentation as something that won't work with the expected delivery dates on NCR's networking solutions. Since SYNTAX advertises that it's SMB server is NFS compatible this may be the solution for us. Anyone else care to share anything with me about similar situations? I would prefer email, but if enough response is generated, I will post a summary here. Thanks. -Bob -- Robert A. Monio National Information Services, Inc. "Market Strategy? Business Plan? pnessutt@nis.mn.org This is absurd!" ..uunet!rosevax!nis!pnessutt
howard@cos.com (Howard C. Berkowitz) (02/21/89)
In article <1689@umbc3.UMBC.EDU>, digennar@umbc3.UMBC.EDU (Mr. Jerry DiGennaro) writes: > > The firm I work for has two and ahalf buildings running LattisNet at > this time.... > . The only major problems has been with the initial > wiring. Phone people think any four wires will make a four wire > circuit. LattisNet needs a continous twisted pair for the receiving > side and a continous twisted pair for the transmit side. Continous > means from the back of the PC/computer/host all the way back to the > LattisNet concentrator. I wonder if you could clarify if your problems with paired wire are due to a loss of pair sense, or are you saying that you have had problems with wire closet punchdowns? In a pair sense problem, some interpretation of installation practice causes two wires which start out in one pair to be assigned to two different pairs. For example, there is a nonobvious point in DECconnect async jack installation instructions, which has a good reason in a pure DEC environment (essentially avoiding the need for a separate null modem with DECservers), which will cause certain "paired" wires at the office wall jack not to appear in the same pair in the wire closet. When this jack was reused for Wangnet (through a dual-coax-to-twisted-pair balun), it simply didn't work, although it did when rewiring was done to keep all pair wires together. In the case above, the twisted pairs, in the eventually working condition, went through 2-3 wire closets containing 66-type punchdowns and/or RJ45 patch panels. As long as the pairs were kept together, everything worked. In the original comment about Lattisnet, would the above wiring closet arrangement be a problem, assuming punchdowns, etc., are done properly? Well-installed punchdowns should seem like the twisted pair, but Murphy doesn't always allow that! Does Lattisnet really need a continuous wire which does not go through wire closets (with the caveat that these closets are wired by competent data installers as opposed to general voice people)? -- howard@cos.com OR {uunet, decuac, sun!sundc, hadron, hqda-ai}!cos!howard (703) 883-2812 [W] (703) 998-5017 [H] DISCLAIMER: Opinions expressed are not necessarily those of the Corporation for Open Systems, its members, or any standards body.
jc58+@andrew.cmu.edu (Johnny J. Chin) (02/22/89)
The makers of LattisNet are SynOptics: (I'm not affiliated with them.)
SynOptics Communications, Inc.
501 East Middlefield Road
Mountain View, California 94043-4015
(800) USA - 8023
I believe that the phone number is correct. I know that the address is.
-- J. Chin (a.k.a. Computer Dr.)
xxxxxxxxxx
xxx xxx xx ------------------ Carnegie Mellon University ------------------
xxx xxx xx 4730 Centre Ave. #412 ARPAnet: Johnny.J.Chin@andrew.cmu.edu
xxxxxxxxxxxx Pittsburgh, PA 15213 BITnet: jc58@andrew.BITNET
x xxxxxxxx x (412) 268-8936 UUCP: ...!harvard!andrew.cmu.edu!jc58
xx xx ----------------------------------------------------------------
xxxxxxxxxx Smile! -- Mr. HappyWOWface -- (got this from the network)
Disclaimer: The views expressed herein are STRICTLY my own, and not CMU's.jsa00564@uxa.cso.uiuc.edu (02/24/89)
howard@cos.com writes: >..the twisted pairs, in the eventually working condition, >went through 2-3 wire closets containing 66-type punchdowns and/or >RJ45 patch panels. As long as the pairs were kept together, everything >worked.. > >In the original comment about Lattisnet, would the above wiring close >arrangement be a problem, assuming puchdowns, etc., are done properly> [stuff deleted] >Does Lattisnet really need a continuous wire which does not go through >wire closets (with the caveat that these closets are wire by competent >data installers as opposed to general voice people. a couple comments I can make here..... 1. No, Lattisnet does not require a "straight-through" cable with no punches. BUT, the punchdowns are the source for problems, and should be minimized whenever possible (without bypassing good practice) 2. At my installation, we have workstations connected 360 feet away, and they go through 3 wire closets by the time the wire gets to the concentrator (hub of the star topology). 3. As I stated before, I was able to simplify the installation and conform to "telephone standards" by converting the Lattisnet RJ45 for use with RJ-11 jacks. Since Lattisnet only uses 4 wires (2-pair) the other 2-pairs in an RJ45 are unused. I had Black Box construct cables to make the switch. ** Things to check if a workstation can't connect to a file server ** This is my basic checklist for diagnosing Lattisnet cable problems, without the use of Sniffers, etc.. (assuming the PC/NIC are good) -- If the workstation is plugged in, and the wire is the correct one, a green light will appear on the NIC and the Concentrator. This light shows ONLY that at least one wire of each pair is connected. Getting the green light (Link Status) is only half the battle. -- Be sure all strapping/jumper wire used in the phone closets are true solid-core, twisted pair. No exceptions. -- Check for obvious bad punchdowns. Repunch all with a SHARP TOOL! -- If it still doesn't work, re-punch each block, starting at the furthest connection from the PC (where it's most sensitive). -- Tighten all clips on 66-type blocks. Replace any that don't seem to fit snug. -- Lastly, the workstation, if it still doesn't work, is probably more than 360 cable-feet from the concentrator. Install a Link extender, or find another wire path.... (I know this is vague) By the way, my conversion to RJ11 jacks allowed me to use our "general voice people" to install Lattisnet nodes, without any special training. The only portion of the wire installation that is different from a phone extension is that Lattisnet's host module (MOD-407) block skips one pair between nodes... E-mail any questions... --------------------------------------------------------------------------- Jeff Autor jsa00564@uxa.cso.uiuc.edu "Help me! I'm an engineering student!" at the University of Illinois (also affiliated with Valmont Electric, a subsidiary of Valmont Industries) The opinions are my own and not necessarily those of Valmont Electric. ----------------------------------------------------------------------------
digennar@umbc3.UMBC.EDU (Mr. Jerry DiGennaro) (02/25/89)
In article <15364@cos.com> howard@cos.com (Howard C. Berkowitz) writes: >In article <1689@umbc3.UMBC.EDU>, digennar@umbc3.UMBC.EDU (Mr. Jerry DiGennaro) writes: >> . The only major problems has been with the initial >> wiring. Phone people think any four wires will make a four wire >> circuit. LattisNet needs a continous twisted pair for the receiving >> side and a continous twisted pair for the transmit side. Continous >> means from the back of the PC/computer/host all the way back to the >> LattisNet concentrator. > > >I wonder if you could clarify if your problems with paired wire >are due to a loss of pair sense, or are you saying that you have had >problems with wire closet punchdowns? > >In a pair sense problem, some interpretation of installation practice >causes two wires which start out in one pair to be assigned to two >different pairs. This is the problem I was mentioning. Synoptics (manufacturers of the LattisNet product) recommends AT&T PDS wiring to include 66 blocks. One does not need solid continuous unshielded twisted pairs running end to end. There is also a recommendation for patch panels to make changes easier. One wires from the workstation to the 66 block in the closet. From the 66 block you go to a patch panel (say with 25 pair cable), from the patch panel, jumpper cables with RJ-45s on them take you to the concentrator. Also the concentrators are linked via fiber optic cables. The maximum length of the UTP wire is some 360 feet or so. (Check the specs for the exact value) Sorry for the confusion with my poor terminology. -- Jerry DiGennaro digennar@umbc3.umbc.edu (301) 266-4150