pcb@cacs.usl.edu (Peter C. Bahrs) (11/28/90)
I have had two netware labs running for 2 and 3 years each, not too much problem. I inherited an old lab to take care of: zenith286 server running a newer version of lantastic, clients are slow 4mhz zenith 100 or 148 pcs with just 2 floppies and 640k. They use a 2mb arcnet network. Problem: (I have posted before) When the clients boot simultaneously, beep,beep,beep what I think is too much contention, clients can't connect and get messages such as error on sector C: (the server). All clients and the server must then be rebooted (corruption!). If I start just a few machines at a time, everything runs fine, even networked wordperfect. What the heck is going on? I am fed up with it. I have never seen a similar problem with netware. Is this common to networking (i.e. so much contention the network or client crashes)? Where is the small degree of fault-tolerant software protocol? Of course our netware's run ethernet...faster than our arcnet. ANYWAY, I have decided to hang a netware server off of the old lab and install one of the arcnet boards in it and try to boot the clients with netware. I hope this will solve the problem. Or will it? In case I am overlooking some 'feature' of lantastic that I have not used, I will apologize now...I am very comfortable with netware though. What do you think? Will my switch help? /*----------- Thanks in advance... --------------------------------------+ | Peter C. Bahrs | | The USL-NASA Project | | Center For Advanced Computer Studies INET: pcb@swamp.cacs.usl.edu | | 2 Rex Street | | University of Southwestern Louisiana ...!uunet!dalsqnt!gator!pcb | | Lafayette, LA 70504 | +-----------------------------------------------------------------------*/
craig@phobos.cis.ksu.edu (Stuart A Craig ) (11/29/90)
pcb@cacs.usl.edu (Peter C. Bahrs) writes: >I have had two netware labs running for 2 and 3 years each, not too >much problem. >I inherited an old lab to take care of: zenith286 server running a >newer version of lantastic, clients are slow 4mhz zenith 100 or 148 pcs >with just 2 floppies and 640k. They use a 2mb arcnet network. >Problem: (I have posted before) When the clients boot simultaneously, >beep,beep,beep what I think is too much contention, clients can't connect >and get messages such as error on sector C: (the server). All clients >and the server must then be rebooted (corruption!). If I start just >a few machines at a time, everything runs fine, even networked wordperfect. >What the heck is going on? I am fed up with it. I have never seen a >similar problem with netware. Is this common to networking (i.e. so >much contention the network or client crashes)? Where is the >small degree of fault-tolerant software protocol? Wow! What an interesting problem. What it sounds like is there are several clients attempting to connect simultaneously to a server that has too few network tasks available (ie. a race condition). Network tasks allow the server to concurrently process several network requests. The more network tasks you specify, the more concurrent requests the server can handle. LANTASTIC generally recommends the number of network tasks be set according to the formula (1/4 * Number_of_clients) + 1 In general, this works well as network requests are usually staggered amoung the clients rather than occuring simultaneously. In this case however, the clients are all the same or VERY similar machines and the connect requests (login) are probably imbedded in the autoexec.bat file, thus all or most of these requests hit the server at the same time and, beep beep beep, race condition. Unless the number of network tasks is equal to the number of clients this is going to happen every time the clients boot simultaneously. The tradeoff is each network task requires a network buffer, and other than at bootup, it is unlikely that all the clients will be making simaltaneous network requests and the formula above would be able to handle the traffic on the server. Consequently, if the amount of memory on the server is not a consideration, then by all means bump up the number of network tasks. Otherwise you might try to fiddle around with the autoexecs so that the connect requests occur at different points on each client. A small time lag would probably do it. And of course there is always the option of not booting everything at once. >Of course our netware's run ethernet...faster than our arcnet. >ANYWAY, I have decided to hang a netware server off of the old lab and >install one of the arcnet boards in it and try to boot the clients with >netware. I hope this will solve the problem. Or will it? >In case I am overlooking some 'feature' of lantastic that I have not used, >I will apologize now...I am very comfortable with netware though. >What do you think? Will my switch help? >/*----------- Thanks in advance... --------------------------------------+ >| Peter C. Bahrs | >| The USL-NASA Project | >| Center For Advanced Computer Studies INET: pcb@swamp.cacs.usl.edu | >| 2 Rex Street | >| University of Southwestern Louisiana ...!uunet!dalsqnt!gator!pcb | >| Lafayette, LA 70504 | >+-----------------------------------------------------------------------*/
35007_321@uwovax.uwo.ca (Gord Nickerson) (11/29/90)
> I inherited an old lab to take care of: zenith286 server running a > newer version of lantastic, clients are slow 4mhz zenith 100 or 148 pcs > with just 2 floppies and 640k. They use a 2mb arcnet network. > If I start just > a few machines at a time, everything runs fine, even networked wordperfect. > > Of course our netware's run ethernet...faster than our arcnet. > Solutions: 1. replace the arcnet with ethernet 2. get version 3 of lantastic 3. buy artisoft netbios driver for lan cards