markat@cyrano.llnl.gov (Mark Tassinari) (06/14/91)
We are in the process of converting a large (4000+ nodes) bridged network to a more manageable (hopefully) subnet topology. The network is divided up into administrative domains that usually have just a single subnet, though there are several that have multiple subnets on a single cable. We are making use of Proxy ARP to ease the transition. After these nets are moved to routers, it it worth going back to each host to change the subnet mask? This can be a big job. I can see no compelling reason to go either way. Changing the subnet mask is the technically correct thing to do. However, for those networks that have multiple subnets on one cable, changing the subnet mask creates additional traffic. If two hosts that are on different subnets (but on the same cable, and possibly sitting next to each other) need to communicate, each packet has to go through the router. Have I missed something obvious? I don't see a good reason to go and change all those subnet masks. What are the advantages of doing it 'right'? Thx. Mark ~~~~~~~~~~~~~ Mark Tassinari markat@cyrano.llnl.gov Lawrence Livermore National Lab (415) 422-4090 PO Box 808 L-408 Livermore, CA 94550
todd@Quotron.COM (Todd Booth) (06/14/91)
markat@cyrano.llnl.gov (Mark Tassinari) writes: >However, for those networks that have >multiple subnets on one cable, changing the subnet mask creates additional >traffic. If two hosts that are on different subnets (but on the same cable, >and possibly sitting next to each other) need to communicate, each packet has >to go through the router. If you must use the correct netmask and have multiple subnets on the same cable, you could add multiple routes to avoid extra overhead. Just add one route for each extra subnet going to the local interface. --todd -- todd (booth) todd@quotron.com 213 302-4368
rexm@lookout.uswest.com (Rex Mammel) (06/20/91)
I agree that the address space savings is valuable. The biggest problem we seem to be facing is a large (15000) population of users/ casual administrators who don't have the understanding to set masks according to specifications or to reserve subnet broadcast addresses/ subnet 0 addresses (ie. not give them out as node numbers.) Perhaps the best solution is to depend on the router for all subnetting and arping, and set every host at class b mask. But there still seems to be some problem with yellow pages, greping for subnets and having to explain the scheme to our users. A modest cost for efficiency, I hope.