dricejb@drilex.UUCP (Craig Jackson drilex1) (01/18/91)
It's looking like it's time to get some real internet addresses around
here, and I have a curiosity about whether to try to get a Class B
and subnet it, vs a bunch of Class Cs.
It seems to me that in choosing between a subnetted class B vs a bunch of
class Cs, the 'cost' will be the same within one's own net. ('cost' being
size of router packets, etc.) The real advantages of a class B only
show up when you connect your net to a larger internet. In addition,
the additional 'costs' of the class Cs are borne by the rest of the
internet, rather than by the owner of the class Cs. (Assuming that
both the class B and the nest of class Cs would have one gateway to
the given internet.)
Am I correct? If I think that the chances on our joining an internet outside
of our company are slim and none, but think I need quite a few nets, is
there any reason to work extra to get a class B?
--
Craig Jackson
dricejb@drilex.dri.mgh.com
{bbn,axiom,redsox,atexnet,ka3ovk}!drilex!{dricej,dricejb}barmar@think.com (Barry Margolin) (01/19/91)
In article <20839@drilex.UUCP> dricejb@drilex.UUCP (Craig Jackson drilex1) writes: >It's looking like it's time to get some real internet addresses around >here, and I have a curiosity about whether to try to get a Class B >and subnet it, vs a bunch of Class Cs. >Am I correct? If I think that the chances on our joining an internet outside >of our company are slim and none, but think I need quite a few nets, is >there any reason to work extra to get a class B? Your description of the routing overhead in the two approaches is correct. An advantage of class C networks over a subnetted class B net is that most TCP/IP software will default the subnet mask automatically if you don't specify it. Of course, if you aren't ever going to connect your network to an outside internet, then you don't even have to get an officially-registered network number; you can even give yourself a class A network if you want! Also, you say "work extra to get a class B"; my understanding is that it should be easier to get a single class B network number than a bunch of class C's, because they assume that you'll be connecting to the Internet and they don't want to bloat the routing tables. We were given a class B network last year so that we could consolidate about a half-dozen class C networks. But if you want to leave the possibility open, and don't want to have to reassign addresses to all your hosts (which, by the way, is tedious, but not really as hard as it is often made out to be -- due to the aforementioned consolidation, over the last six months we've changed the addresses of most hosts on our net (about 400 Unix systems and at least 60 Macs), and changed many of the Suns twice), then getting official network numbers is a good idea. Subnetting a class B makes some network management and administration tasks easier. You get to decide how many bits to allocate to subnet number; the NIC probably won't give you 254 class C's, but you can probably use eight bits of a class B as the subnet number. Since you control the subnet numbers, you can attach semantics; for instance, on our network, subnets 1-15 contain hosts directly accessible from the outside networks, subnets 16-127 are for administrative and research computers, and subnets 128-254 are for development computers, and our router packet filters can easily recognize these blocks of addresses using bitmasks. This also makes it easier to recognize and remember your network numbers, as they are effectively only two or three digits long (the two-octet class B prefix will quickly become automatic). On the other hand, if you get class C nets, the network numbers will be pretty arbitrary and may not even be consecutive (especially if you get more networks later on). -- Barry Margolin, Thinking Machines Corp. barmar@think.com {uunet,harvard}!think!barmar
reschly@BRL.MIL ("Robert J. Reschly Jr.") (01/20/91)
Craig,
Barry covered things pretty well. The only other issue which comes
to mind is that many IPs derived from BSD code allow you to set a
"subnets are local" flag. Setting this flag causes the code to generate
MTU sized datagrams rather than 576 octet datagrams for all subnet
destinations. In most subnetted environments, this is a significant
win.
Later,
Bob
--------
IP: reschly@BRL.MIL UUCP: ...!{{cmcl2,nlm-mcs,husc6}!adm,smoke}!reschly
U.S. Army Ballistic Research Lab. / Systems Eng. & Concepts Analysis Div. /
Networking & Systems Dev. Team / Aberdeen Proving Grounds, MD 21005-5066 /
ATTN: SLCBR-SE-A (Reschly) // (301) 278-6808 FAX:-5075 DSN:298-
**** For a good time, call: (303) 499-7111. Seriously! ****