---------------------------------------- CONNECTION PROBLEMS UPDATE: 23 JUNE 1996 ---------------------------------------- It's time for a progress report on the connection problems some of you have been reporting. This is going to be fairly long, but we encourage you to read it so you will know what we have done so far to diagnose the problems and what still remains to be done. Let's begin by reviewing the problems: Ring with no answer ------------------- If your modem dials and you hear an endless ringing with no answer, then either you have hit a BCPL modem that refuses to answer or the phone line is broken between Bell Atlantic's Towson central office and our computer room. This has been a recurring pr oblem. With one exception involving a dead modem, all instances of "Ring - No Answer" have proven to be caused by dead phone lines. A major contributor to the problem is the fact that the cabling between the Towson CO and our computer room is a patchwork involving many splices and connections, any of which can go bad due to age or human error. There is now a fiber optic cable running from the Towson CO directly to our computer room which we already use to bring in our high-speed T1 lines. We have met with representatives of Bell Atlantic and Cisco Systems to discuss the feasibility of bringing our modem lines in over the fiber as well. One advantage of this is that the fiber optic cable is a continuous run from the Towson CO to BCPL's computer room, so there is less likelihood of failure. Another advantage is almost complete freedom from induced line noise. Answer with no modem negotiation -------------------------------- If your call is answered but you don't hear the usual shrieking noises from your modem's speaker, then your modem and ours have failed to negotiate the connection. The negotiation phase is where the two modems agree on the connection speed they will use, and on the error correction and data compression methods to be used. If this negotiation doesn't occur, then the connection drops. The puzzling thing about this scenario is that some callers experience it fairly often, while others never experience it at all. This type of problem can be caused by a defective or incorrectly configured modem at either end of the connection. It can also be caused by a connection that drops before the negotiation can take place. Some modem experts say it can also be caused by a switching failure somewhere along the route between you and BCPL. In this case your mode m may think the call has been answered even though it hasn't really gotten through to a BCPL modem at all. One sign of this is a series of clicking noises heard over your modem's speaker instead of the usual shrieking sounds of modem-to-modem negotiation. On July 21 Bell Atlantic will be replacing all the switches in the Towson CO with a more modern type, which may or may not help. Username prompt, password prompt, or menu fail to appear -------------------------------------------------------- In this scenario you are doing a manual login but get stuck somewhere along the way. The "Username:" or "Password:" prompt may fail to appear or may not accept your input. Or after entering your username and password, the usual "PPP, UNIX shell, Exit & Disconnect" menu fails to appear or only partially appears. If you are using a scripted login, any of these will cause the login sequence to get stuck part way through the script. This is a difficult problem to troubleshoot because there are so many possible causes. It may be that the connection dropped, stopping the login process. It may be a failure of our modem or yours to send or receive data. It may be a failure of our terminal server to send out the prompts and/or the menu, or to accept your input. It could be software failure at the your end. Or it could be something else entirely. This is another scenario that some callers experience often, but other callers never experience at all. Connection made and working, but drops -------------------------------------- This is probably the most widely reported problem. In this scenario your modem establishes a connection with BCPL's. The two modems negotiate the connection successfully (those shrieking sounds). You are able to log in successfully. You start doing whatever it is you do on the Internet and all appears to be going well, but then the connection drops well before your three-hour session limit expires. This is another difficult problem to troubleshoot. One of the realities of modeming is that the higher the connection speed, the more fragile the connection becomes. At 14.4 kbps and 28.8 kbps there are many things that can interrupt a connection, and many of those things wouldn't have phased your old 2400 baud modem at all. Unexpected disconnects will occur occasionally at high modem speeds. It's a fact of life in the high-speed modem world. However, frequent disconnects should not occur. We are concerned about the frequency of disconnects reported by some users, and especially about the "I got disconnected after about two minutes" reports that we hear so often. That "two minutes" theme is fairly common in the reports. It sounds almost like a timer somewhere ticks off two minutes and then cuts you off. This is puzzling because there is no timer setting in our modems or in our terminal servers that can cut you off after two minutes. Also puzzling is the fact that this is another scenario experienced by some callers fairly often, but never by others. What we have done ----------------- Here is what we have done so far in our efforts to solve these problems: Asked callers suffering from dropped PPP connections to report the IP addresses they had during those connections. This allowed us to determine which modem they were connected to at the time. The expectation was that this would narrow the problem down to a few faulty modems, but in fact it showed that a dropped connection can occur no matter which of our modems you connect to. Flash upgraded our MultiTech MT2834BR rack-mount modems to the latest PROM version (2.13). Previously they used a variety of PROM versions ranging from 2.08A to 2.12, depending on the age of the individual modem. They have also been checked several times to verify that every one is configured according to MultiTech's recommendations. Verified that our U.S. Robotics Quad Modems, which are in a self-contained unit called a "Cisco AS5100 Access Server", are configured according to USR's and Cisco's recommendations. Watched all modems for many hours to verify that each one is capable of answering incoming calls and that each one is capable of establishing and maintaining connections. Verified that our Cisco 2511 terminal servers are all configured according to Cisco's recommendations. Had Bell Atlantic test all of our incoming phone lines for noise level and other problems. A few minor problems were found and repaired, but nothing likely to be causing serious connection problems. Worked with callers to confirm that their modems and software are configured correctly. This is difficult due to the wide variety of modems and software being used by our callers, but our Help Desk staff are gradually becoming familiar with a wide range of products. Monitored "Ring - No Answer" problems on a continuing basis and worked with Bell Atlantic to correct them. What next? ---------- We will continue to work with MultiTech, U.S. Robotics, and Cisco to resolve the connection problems, but there is probably little to be gained by the same people going over the same modem and terminal server configurations over and over again. Therefore we are bringing in an outside consultant in the hope that he will be able to spot something we and the equipment manufacturers have missed. We will move our incoming lines to the new fiber optic cable if it proves economically feasible to do so. This will ensure noise-free lines between BCPL and the Towson CO, and will greatly reduce the likelihood of lines randomly going dead as they have in the past. Our Help Desk staff will continue to work with callers reporting connection problems. In addition they will assume an information gathering role to help us determine why some callers have connection problems and others do not. There has to be a reason for this. If we can discover that reason it may point toward a solution. Conclusion ---------- We know that these connection problems have been extremely frustrating for those of you who have experienced them, and we appreciate your patience while we diagnose, troubleshoot, and try new solutions. We have no doubt that the correct solution will eventually be found, and we hope you will continue to be patient while we continue to work toward that end.