NATIONAL RADIO ASTRONOMY OBSERVATORY Green Bank, WV MEMORANDUM September 9, 1997 TO: E. Childers, R. Fisher, J. Ford, R. Lacasse, R. Norrod, M. McKinnon, C. Niday, T. Weadon FROM: R. Lacasse SUBJ: Minutes of Interference Protection Group 9/9/97 Meeting In attendance: Childers, Fisher, Ford, Lacasse, McKinnon, and Niday, Norrod, Weadon The main purpose of the meeting was to discuss NRAO's role in the impending installation of a LAN in the Green Bank Elementary/Middle School. As background for this meeting, John Ford wrote a memo summarizing options and costs. For the record, this memo is included below, between the asterisks. Minutes of the meeting follow. **************************************************************** Date: Tue, 2 Sep 1997 17:40:30 -0400 From: John Ford To: tweadon@sadira.gb.nrao.edu Cc: rfisher@hyades.gb.nrao.edu, rnorrod@NRAO.EDU, rlacasse@NRAO.EDU, tweadon@NRAO.EDU, mmckinnon@NRAO.EDU, jlockman@NRAO.EDU Subject: Re: LAN's There are several things to consider in the planning of the network for the GB Elementary/Middle School. One is current needs. Another is the interference problem. Yet a third is the future plans for the network and P.C.s at the school. Finally, the matter of funding. Current needs: There are about 15 computers to be networked currently, including 8 Power Macintosh computers. These computers all have 10BaseT network adapters in them. Obviously, the cheapest way to connect them up is to put unmanaged hubs and UTP cable in place. I have designed the network so that there is a 100 Mbit connection to the server, and each disparate workgroup, or group of computers, shares a 10 Mbit segment of the switch. the topology is then: server <-100Mbit->Switch+<--WG1 (unmanaged hub) +<-- WG2 (unmanaged hub) +<-- WG3 (unmanaged hub) +<-- internet router Crudely costing this out we have: 10BaseT: Qty Cost Description 3 300.00 Unmanaged hubs x8 ports 1 759.00 Bay Networks Netgear switch (10/100 baseT Ethernet Switch 8 10 Mbitports, 2 100Mbit ports) 25 30.00 crimp-on connectors 1000 ft 375.00 cable (non-plenum) 15 40.00 plastic wall boxes and outlets 15 150.00 plastic wiring channel For a shielded twisted pair system: Add $150.00 for Cable and Connectors Total: $1654.00 (UTP) $1805.00 (STP) Coaxial cabling: 1000 ft 150.00 cable 1 3247.00 10/100 ethernet switch (100 Base FX and 10 base2) 10 29.00 plastic outlet boxes 15 60.00 T connectors & terminators 10 100.00 plastic wiring channel 15 1500.00 Coax Ethernet cards $5146.00 Fiber: 1000 ft 600.00 fiber 60 420.00 Connectors 1 3247.00 10/100 ethernet switch (100 Base FX and 10 base2 ) 1 3050.00 16 port fiber hub 15 29.00 plastic outlet boxes 10 100.00 plastic wiring channel 15 2500.00 Fiber Ethernet cards $9946.00 In addition, the future add-ons are likely to be Fast Ethernet, which requires either fiber or Twisted pair. The plan is to have 2 computers in each classroom in the school by 1999. At that point, there will be about 40 pentium class PC`s at the school networked together. Each drop for fiber costs about $500.00. Each drop on UTP costs about $100.00. Coax is also about $100.00 per drop. The school has 10k to spend on a network and a server, including software. I figure that at least 6k needs to be spent on the server, leaving less than 4k for the network. I guess my first choice would be to go with coax. It`s not a sexy as fiber, but eminently more affordable. If we had the money, it might be worth making the school into a showcase for interference supression using fiber. It may be an interesting experiment to put in the Shielded Twisted Pair, and try it out. The cost to install it is low, and if it works to our satisfaction, fine. If not, we will have to go back and retrofit the system with coax or fiber. In any case, the school needs to get started on soon, so we don`t have too much time to ponder this. John **************************************************************** Fisher: The possibility of interference to the GBT from the school installation is clear. The LAN is only part of the installation, and it is not clear how much interference it will generate compared to the computers and other electronics at the school. We need to make it clear to the school that interference is important to our operations. Ford: A twisted pair network will probably cause trouble. Any sort of damage to the cable, such as that caused by tripping over the cable, could increase radiation from the shielded pair. Are the computers much worse than the LAN? The school has a lot of computers and is getting 8 more Power Macs. LAN's are rated Class A, which requires no filtering, but the ethernet spec imposes limits on the wave shape, limiting the edge rates. Childers: Most people worry about other things interfering with their network. A balanced line and constant impedance are the key to suppressing interference from the network connections. Ford: Found some info on the web from a European lab that claimed an additional 20 to 30 db of shielding from shielded twisted pair (vs unshielded twisted pair). Twisted pairs are great in theory, but not in practice. Childers: The termination of shields is difficult. Good contact with the backshell is tricky, and patch panels can also be tricky. Shielded twisted pair installations are becoming popular in Europe, where regulations are more stringent. Ford: Class A equipment is not tested by the government; it is self-certified. Is coax better than shielded twisted pair? Ford: The school's goal is to have 2 computers in each classroom by 1999, and to have all computers on a LAN. Weadon: There are legal means we can use if necessary to reduce the interference from the school. Childers: There is a state law that can be used in our favor. Also the use of computers and LAN in the community will certainly grow. Fisher: In this case, we need to take on some of the responsibility ourselves. Lacasse: Also the school is the closest and largest potential source of interference locally. Fisher: We need to think about which solution would be most cost effective in the long run, taking into consideration our time investment. Ford: 10BaseT will become 100BaseT in a couple of years. There is no path to 100BaseX from coax, but the 10BaseT can become 100BaseT simply by changing the equipment on the ends of the cable. (This raises the signalling rate to only 33MHz as the signal is split over more pairs). Childers: Running 100Mb ethernet over unshielded twisted pairs is not practical, because the distances between repeaters gets tool small. Norrod: Why does the shielding effectiveness go down with time? Childers: Mismatches get worse at the connectors due to handling, and cables can get damaged. A hybrid network, like we have at NRAO, consisting of a fiber backbone and coax rings works well. However, coax does have the disadvantage that the entire ring can be brought down by someone disconnecting a cable; that can be a problem in places where people are not technologically sophisticated. McKinn: The installation should be simple enough that the school can do a lot of the maintenance itself. Ford: All systems are fairly reliable. Twisted pair in a star configuration is probably simpler to troubleshoot that a coax ring. Weadon: What does the estimate of $10K for fiber cover? Ford: It connects 15 computers. The cost is roughly $500 per computer for fiber. Since the school's goal is 48 computers, another 33 computers will be added in the near future, requiring an additional $17K for a fiber network. The additional cost of fiber would possibly slow the implementation of the plan. Coax is becoming obsolete. Shielded twisted pair appears to be the most viable option. Weadon: We are opting out of the fiber route because of cost. Could management find this money and make it available? Lacasse: We need a decision very quickly, so it is unlikely we could get a commitment from management in the required time frame. Fisher: It sounds as if we should recommend shielded twisted pair as the best option for now. Technically, it seems reasonable, and it is also enough of a departure from common practice to send a message about the importance of interference at NRAO. Lacasse: It would seem a shame to not make a quick measurement of two computers, with and without a LAN connecting them, just in case there are surprises. Ford: We could intercept the first two MACs that come in. ?: What is the longest run anticipated at the school? Ford: About 150'. Fisher: Then we should use a similar length of cable in our test. Childers: Some PC's we already have similar LAN connections and could be used for these tests. Other items: Call outs for interference: None Status of spectrum analyzer purchase: Niday: There are three viable options, one from HP, one from Advantest, and one from Anritsu. Have one outstanding question about the Advantest unit, about averaging. Lacasse: We should let Omar get best and final offers on these units and proceed with the purchase. Niday: Initial quotes ranged from 26 to 32K. Weadon: The total budget for the IPG for this year is 35K. By next meeting an estimate of remaining funds should be available. Meeting times: Weadon: Having the meeting right before the Coordination meeting is inconvenient for me. Fisher : Others having problems? All: No. Fisher: Let's discuss this item next meeting. What about meeting frequency? Lacasse: Two weeks feels about right for the infant stage of the IPG. We might want to switch to once per month once things become more routine, but I don't think we're there yet. All: General agreement.