remaker@pepsi.amd.com (03/01/90)
I have a problem. Here at AMD we have a well administered highly heterogenous TCP/IP and DECNET network. Recently we have come under a great deal of pressure to support novell LANs on the network. Now I know that these protocols can exist on the same backbone and be routed by our cisco routers. I hope to get advice on the following: 1) How can I centrally administer these nodes? Since the hardware address is used as the network address, does this imply that I have to keep a database of the hardware addresses? 2) How can I monitor traffic from a Unix platform? etherfind is no help and etherview is kind of ugly since I'd have to sit and disect individuall type 8137 packets. 3) What kind of diagnostic utilities (on a Unix platform) are available? For IP I have ping, etherfind and traffic on my Sun. I can log into a VMS machine and do NCP monitoring for DECnet (or I can use an Ultrix/VMS combined station). However, no such diagnostic utilities exist for Novell. I do NOT want a DOS station in my cube only to monitor Novell. I want centralized troubleshooting at one point. 4) How can we do traffic accounting for bandwidth management? I want to keep the broadcasts as localized as I can IF I have to support IPX, and I want to know WHO is producing the most traffic, and how to best localize it. This is especially a concern in keeping our long lines free of extraneous traffic from careless mounts (a current problem with NFS). 5) Tell me any advice and experience you all have putting Novell, 3Com, Banyan, and LAN manager based machines on your TCP/IP backbone. I appreciate any help. I'm highly concerned, since IPX was designed as a protocol for small LANs, I fear it will grossly misbehave in a WAN environment. I also fear that the traffic among these LANs may be prohibitive. But what I fear the most is the inability to adequately track, troubleshoot and document these nodes, since we will not be assigning addresses for them, and it would be hard to track down illegal installations. We run a tight ship here and need 24 x 7 availability, and we simply cannot afford chaos on out network. Thanks again for all of your help. I will post a summary if sufficeient interest exists. P.S. Vendors like Novell & 3Com were baffled by this request at NetWorld 90. Why would you want that? Just make your TCP/IP hosts speak IPX! Sigh. 8-(. I turn to the net for the definitive answer 8-). Phillip A. Remaker A.M.D. M/S 167 P.O. Box 3453 Sunnyvale, CA 94088-3000 remaker@amdcad.amd.com Cutting Edge Networking...Close to the Jugular... "It's only work if someone makes you do it." -Calvin
haas@cs.utah.edu (Walt Haas) (03/01/90)
In article <29319@amdcad.AMD.COM> remaker@pepsi.amd.com () writes: >I have a problem. Here at AMD we have a well administered highly heterogenous >TCP/IP and DECNET network. Recently we have come under a great deal of pressure >to support novell LANs on the network. Now I know that these protocols can >exist on the same backbone and be routed by our cisco routers. I hope to get >advice on the following: > >1) How can I centrally administer these nodes? Since the hardware address > is used as the network address, does this imply that I have to keep a > database of the hardware addresses? Why do you need this information? The routers keep track of addresses quite nicely all by themselves. >2) How can I monitor traffic from a Unix platform? etherfind is no help and > etherview is kind of ugly since I'd have to sit and disect individuall type > 8137 packets. I haven't found a method, if you find one I'd be interested. The Network General Sniffer does a nice job of translating to English. >3) What kind of diagnostic utilities (on a Unix platform) are available? > ...I want centralized troubleshooting at one point. See answer to 2) above. >4) How can we do traffic accounting for bandwidth management? I want to > keep the broadcasts as localized as I can IF I have to support IPX... The ciscos do a good job of blocking broadcasts. > and I want to know WHO is producing the most traffic, and how to best > localize it. > This is especially a concern in keeping our long lines free of extraneous > traffic from careless mounts (a current problem with NFS). Designing for good localization is a bear. Somebody at Network General said they are bringing out a new product that will help with this. I immediately started salivating and demanded to be a beta site :-). >5) Tell me any advice and experience you all have putting Novell, 3Com, > Banyan, and LAN manager based machines on your TCP/IP backbone. As of 8.0(13) the cisco code seems to be fine. We route Novell just like IP and DECnet with it. Sorry, no experience with those others. > I appreciate any help. I'm highly concerned, since IPX was designed as a > protocol for small LANs, I fear it will grossly misbehave in a WAN > environment. Novell just implemented XNS, it shows up that way on the Sniffer. >I also fear that the traffic among these LANs may be prohibitive. We haven't had any major problem yet, with our 35 Novell fileservers. > But what I fear the most is the inability to adequately track, troubleshoot > and document these nodes, since we will not be assigning addresses for them, > and it would be hard to track down illegal installations. The ciscos will help you localize who is where. > We run a tight ship here and need 24 x 7 availability, and we simply cannot > afford chaos on our network. We have the same goals. We accomplish them by letting the routers do 99.9% of the work. Of course our network may be smaller than yours, I don't know. In general I don't think you would gain anything by forcing users to use TCP/IP rather than Novell, at least until your network grows so large that the routing tables become prohibitive in size (we may be approaching that now with AppleTalk). Cheers -- Walt
morris@jade.jpl.nasa.gov (Mike Morris) (03/02/90)
haas@cs.utah.edu (Walt Haas) writes: >In article <29319@amdcad.AMD.COM> remaker@pepsi.amd.com () writes: >>I have a problem. Here at AMD we have a well administered highly heterogenous >>TCP/IP and DECNET network. Recently we have come under a great deal of pressure >>to support novell LANs on the network. Now I know that these protocols can >>exist on the same backbone and be routed by our cisco routers. I hope to get >>advice on the following: [...edited...] >>2) How can I monitor traffic from a Unix platform? etherfind is no help and >> etherview is kind of ugly since I'd have to sit and disect individuall type >> 8137 packets. > >I haven't found a method, if you find one I'd be interested. The Network >General Sniffer does a nice job of translating to English. He's not the only one that would be interested... >> I appreciate any help. I'm highly concerned, since IPX was designed as a >> protocol for small LANs, I fear it will grossly misbehave in a WAN >> environment. > >Novell just implemented XNS, it shows up that way on the Sniffer. We're using it - Filenet thicknet into one card, one IBM 4/16mb token ring card talk to a client ring, another to a backbone to 6 more client servers. Only problem is that the Novell server won't bridge the XNS. i.e. I can login to the Filenet from a ethernet workstation attached to the hose just fine. But the exact same station with a token ring card hooked to either ring can't find the filenet server.... with or without source routing... BTW, don't flame me for inaccurate names. 5 months ago I did't know anything about Novell - and I'm not the guy who has to make this mess work. I'm just trying to get the gateways working, write a little Windows code (anybody got recommendations on a dBase library for MS C? - final target is Clipper-generated & maintained files), and beat 250-odd AT&T 386sx's into submission - several packages (including Windows 386) don't like their flavor of 800*600 VGA... Mike Morris Internet: Morris@Jade.JPL.NASA.gov Misslenet: 34.12 N, 118.02 W #Include quote.cute.standard Bellnet: 818-447-7052 #Include disclaimer.standard Radionet: WA6ILQ