HOT-WIRED, Part One

by Wayne M. Krakau - Chicago Computer Guide, April 1994 - NewsWare, January 1996

They are so common and so transparent to our daily lives that we ignore them. But, they are so vital to the functioning of our systems that even the smallest problem can disrupt an entire network. They are wiring systems, more properly known as cable plants. Here are some real gems that I have encountered.

I was called in by another LAN reseller to take a look at a 10Base-T (Ethernet over unshielded twisted pair cable) LAN at his client's office. They needed to get several new workstations connected to the LAN and have a server-based tape drive installed on their new network. In addition, they were having problems keeping the only existing workstation connected to the network.

The existing workstation was first. I checked the shell software (that talks to the network) and updated it to the latest version. Then I checked the configuration of the NIC (network interface card). It's always a good idea to see if a NIC is taking up some resource (IRQ, I/O address, high RAM address, DMA) that another device is using. This is one of the most common ways to get a network to act "funny". Reconfiguring the NIC and running tests against it didn't help.

Since the client was a telephone equipment and cabling company, and I was nearly surrounded by professional cablers, I didn't initially check the cabling. When the workstation kept losing its network connection in the middle of testing the NIC, I decided that cabling could be an issue. That's when I found out that the jumpers (terminated cables that go between things) that connected the workstations to the network sockets in the wall were made from what's commonly known as silver satin cable. In English, that's the same wire that goes from a phone to the wall socket in homes! It is flat - that is it has no twists. The IEEE 10Base-T specifications call for unshielded TWISTED pair (called UTP) cable! Since Ethernet is inherently a party-line (as in rural communities, not as in 900 numbers) where every signal is heard by all, even one piece of silver satin can bring chaos to an entire network. This place had a silver satin jumper at every workstation! No wonder they couldn't keep their first workstation connected.

When I explained the standard to the manager (a non-techie), he reacted rather strangely. This made me suspicious, so I asked about the wire in the walls. He explained that since unshielded cable was good, his people had told him that SHIELDED cable would be even better! I re-explained the standard, but I could tell he didn't really believe me. All of his cabling people couldn't be wrong!

Then I asked to see their wiring closets. I confirmed that the wire coming out of the walls was shielded twisted pair (STP), not UTP. I also found that the cables going from the punch-down blocks (where raw wires are connected) to the patch panels (a kind of intermediate box of sockets similar in principle to an old-fashioned phone switchboard) and from the patch panels to the concentrators were also silver satin. Just for bonus points, the panels used the center four pins of RJ-12 6-conductor home-phone style sockets rather than the standard specified 8-conductor RJ-45 sockets.

Between the incorrect wire and the mismatching impedance caused by mixing different types of wire, a linked set of rusty coat hangers could probably transmit a better signal! No wonder they had problems.

I then attempted to get several more workstations connected to the network, but was unable to get more than three up at once. The cable plant just couldn't handle any more. Luckily, the tape installation worked perfectly, so I had something to validate my competence.

In an additional effort to establish that I wasn't making up standards, I called my cabling authority, Bruce Kahn of Telnet Communications Consulting, Inc. (708-215-0003), a local cabling company, and confirmed my findings. Finally, I faxed the client a list of Ethernet specifications that I had compiled for the seminars that I occasionally teach.

Even before I left, however, I got some indication that I was convincing management that their cable plant had problems. I heard the distinctive sounds of one of the technical managers being interrogated - very loudly. Eventually, the company completely rewired the whole place. That solved their problems.

In another situation, I was called in directly by the system administrator of a LAN that occupied one floor of a fairly new high-rise building. We got together to discuss his expansion plans for the network. The client needed more workstations on their 10Base-T LAN, some of them MACs. In addition, the administrator was aware that due to his lack of training, the LAN was probably not optimized. Finally, he told me that he had some problems keeping workstations attached to the network.

The first thing that I noticed was that his recent Netware upgrade was to Netware V2.2, not the then current Netware V3.11. As I have stated before in this column, I consider that a dirty trick and grossly unethical. Resellers will dangle the small difference in price between the two versions in front of clients, but neglect to tell them that V2.2 will cost them thousands extra in initial installation costs and ongoing maintenance. On top of that, it is much slower and much more difficult for the system administrator to manage on a day-to-day basis.

Next, I noticed that they were using the old NETX/IPX twins rather than the ODI shell. The NETX/IPX combo was declared dead (and rightfully so) by Novell as of November of 1991, so I suggested that an upgrade was in order.

Then I ran TXD, Thomas-Conrad Corporation's (800-332-8683) diagnostic program. Just running the diagnostic program was enough to blow the workstation right off the network! I started to examine the wiring system. I found that one of the Thinnet coaxial cables used to connect the concentrators was physically damaged and two others looked to be on their last legs. The administrator immediately agreed to order replacements for all three.

The administrator then pointed out that several of the individual sockets in the concentrators were inoperable. Since they were using bargain-basement concentrators, I suggested that they replace all of them while they still had time. The administrator didn't have the authority to authorize that purchase.

Next, I inspected the cable from the wall to the punch-down blocks. It did not look like any data cable that I had seen before. While it was definitely UTP, it just didn't look right.

I traced the cable from the punch-down block to the patch panel and found that the patch panel end was terminated in an old 50-pin telco Amphenol socket. This socket takes a bundle of wires within one sheath and spreads them out into 50 separate contacts. While I have heard of people trying this trick with 10Base-T before, I noted that this was a violation of standards. Putting multiple high-speed data wires together within that sheathing and especially inside the actual Amphenol connector was asking for trouble.

The front connectors on the patch panel blocks were RJ-12s, not RJ-45s - here we go again. The jumpers from the patch panels to the concentrators and, as it later turned out, from the wall sockets to the workstations were all the funny looking UTP that I had found coming out of the wall.

I got out my LANcat, a test instrument programmed with the IEEE standards for the cables used with all of the common networking systems made by Datacom Technologies, Inc. (800-468-5557) and also sold under the Fluke brand name. I used it to test the cables running from various workstations back to the concentrators. None of the cable passed any of the tests at the Level 3 standard. Levels are a measure of the grade of cable ranging from 1 to 5, with 5 being the highest, not to be confused with "Type", IBM's way of designating different kinds of cable. Though we usually install at least Level 4 for use with 10Base-T to provide more leeway, and many clients now demand Level 5 to allow for later upgrades to high-speed networking, the actual standard specifies a minimum of Level 3 UTP. Failure in the Level 3 tests indicated that the cable plant was useless for 10Base-T.

I took down the numbers on the cables and connectors that I was inspecting and read them to Bruce. He later confirmed that the cables, punch-down blocks, patch panels, and wall sockets were all LEVEL 1! This information was so surprising that I had him send one of his crew down to personally check the cable plant and do more testing. The results corroborated my initial findings.

This was a double surprise. First, Level 1 is basically very low grade obsolete voice-only phone cable. Levels 1 and 2 had not been stocked by most cable distributors since 1990 when the price of Level 3 dropped. This means that not only did the cabling company screw their client (the only appropriate word for it), but they must have gone out of their way to do it! Perhaps they had a left over spool that they wanted to get rid of, so they stuck this client with it. Either that, or they special ordered a spool just for the occasion.

The second surprise was that the cabler was a prominent local company and LANDA (Local Area Network Dealers Association) member who should have known better. Inexperience could not be one of their excuses.

At my suggestion, the client called in the original cablers to get what amounted to warranty coverage. That company claimed that the Level 1 cable was just fine for 10Base-T (and, I presume, that they had conditionally repealed the laws of physics).

They sent one of their technicians to look at the system. He replaced the jumpers at the three most troublesome workstations with Level 5 jumpers and declared the problem fixed. (This is your brain. This is your brain on drugs.) Five minutes after the technician left, the system administrator called me asking for help. All workstations had fallen off the net and none could get back on. (Aha - the dreaded Humpty-Dumpty Syndrome!)

The massive differences in electrical characteristics between the Level 1 and Level 5 cables stopped the network dead in its tracks. This was predictable. Radically different levels of wire put together in this way create a reflection point within the cabling system, scrambling transmissions. I had the administrator replace the new jumpers with the old, thereby clearing up the new problem.

The management of the client company rejected any talk of rewiring, even though the system administrator spent much of his day running from workstation to workstation getting machines back onto the network. They were unable to press home warranty claims with the original installer.

Management did the next best thing. They brought in a mainframe consultant with no cabling or even PC experience to recable all of the workstation and concentrator jumpers, using a reel of cable and a crimping tool from the local Radio Shack! Hey, why didn't I think of that! (Now I know why there is a waiting period for gun purchases.)

Need I tell you the results? The network has deteriorated to the point where the system administrator spends most of his time resurrecting dead workstations. (Picture the Ed Sullivan Show. Picture the plate-spinning balancing act on that show using six plates. Now picture the same act using seventy-five plates.) No further action is being contemplated.

Part Two of the wacky world of wiring will be in next month's issue. I've got to go on a late night rescue. Let's see now, Jolt Cola, Pepto-Bismol, aspirin, TUMs, pizza delivery coupons. Yep, I'm ready to go.