Thanks again !! I got some excellent replies so I'm including them at the
bottom. I want to thank everyone who sent me info. My original question
in a nutshell was:
Help - I'm gonna have a bunch of PPP accounts & someone said I needed to
subnet & the manuals were woefully lacking.
Silly me I looked in the wrong o'reilly book but a bunch of people
pointed me towards the TCP/IP admin book.
I think my network is too simple to start subnetting & proxy arp will
work just fine.
Thanks again!
erin
erin@skivs.ski.org
>From ed@magma.COM Sat Aug 5 10:06:44 1995
To: erin@skivs.ski.org
Subject: Re: subnetting a Class C
Content-Length: 1128
X-Lines: 28
Status: RO
Erin -
A much less expensive way to get to PPP is to use one of our
SBus serial port boards (if you have an SBus machine, you didn't
say). The cheapest is $275, way less than an Annex box.
If you're interested, I'll get sales to send you info.
Regards,
Ed
------------------------------------------------------------
_/ _/ _/ _/_/_/ _/ _/ _/
_/_/ _/_/ _/ _/ _/ _/ _/_/ _/_/ _/ _/
_/ _/_/ _/ _/ _/ _/ _/ _/_/ _/ _/ _/
_/ _/ _/ _/_/_/_/ _/ _/_/_/ _/ _/ _/ _/_/_/_/
_/ _/ _/ _/ _/ _/ _/ _/ _/ _/
_/ _/ _/ _/ _/_/_/ _/ _/ _/ _/
------------------------------------------------------------
Ed Romascan email: ed@magma.com
MAGMA Tel: (619) 457-0750
6725 Mesa Ridge Road #100, Fax: (619) 457-0798
San Diego, CA 92121
------------------------------------------------------------
>From tmornini@parallaxinc.com Sat Aug 5 10:21:47 1995
Subject: Re: subnetting a Class C
To: erin@skivs.ski.org
In-Reply-To: <9508041956.AA23986@skivs.ski.org> from "Erin O'Neill" at Aug 4, 95 12:56:35 pm
X-Mailer: ELM [version 2.4 PL24]
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Length: 678
X-Lines: 13
Status: RO
> So I want to bite the bullet & subnet. I go to my O'reilly book
> on Managing NFS & NIS. It's VERY cryptic (unusual for O'reilly).
> So I go to my admin book (it's Sunos 4.1.2 & I use 4.1.3_u1).
> It's not quite as cryptic as o'reilly but pretty close. Surely
> there exists a paper that gives the nitty gritty details of
> subnetting???
O'Rielly, TCP/IP Administration.
-- Tom Mornini ----------------------------------------------------------
-- Parallax, Inc. ------------------------------------------------------
-- Makers of really cool PIC development tools & the BASIC Stamps ------
-- http://www.parallaxinc.com ftp://ftp.parallaxinc.com/pub --
>From allan@mazama.com Sat Aug 5 15:26:31 1995
To: erin@skivs.ski.org (Erin O'Neill)
Subject: subnetting a Class C
In-Reply-To: <9508041956.AA23986@skivs.ski.org>
References: <9508041956.AA23986@skivs.ski.org>
Content-Length: 922
X-Lines: 21
Status: RO
Rather than subnetting, use proxy arp. I've been using a NetBlazer,
but got an Annex on order. They both do proxy arp.. Using proxy arp
allows the ppp or slit connections to appear to be on your local net.
If you still want to subnet, on you yp master you need to modify the
netmasks map. For example to create 4 subnets from a C you do --
199.101.202 255.255.255.192
You then have the following nets - 199.101.202.0 199.101.202.64
199.101.202.128 199.101.202.192.
This subnets all you unix boxes, but you will have to go configure all
your pc's routers etc with the correct netmasks. You need to check all
your equipment to see if its able to.. It blew my mind that you can't
subnet a C on a Cisco 2500 ethernet port..
-- ---------------------------------------------------------------------- Christoher A. Stewart | allan@eskimo.com Mazama Software Labs. | allan@mazama.com>From jcr@ipl.bwh.harvard.edu Sun Aug 6 11:09:41 1995 To: erin@skivs.ski.org Subject: Re: subnetting a Class C Content-Length: 70 X-Lines: 1 Status: RO
Would you please be so kind as to summarize. (i.e., "me too.") TIA.
>From kevin@uniq.com.au Sun Aug 6 18:26:40 1995 X-Mailer: Mail User's Shell (7.1.2 7/11/90) To: erin@skivs.ski.org (Erin O'Neill) Subject: Re: subnetting a Class C Content-Length: 1121 X-Lines: 24 Status: RO
[ Regarding "subnetting a Class C", erin@skivs.ski.org writes on Aug 4: ]
> All I wanted to do was give my users PPP connections & I feel like > I've entered a dark maze! I'm trying to get the Powers > that Be to replace my ALM boards (which are much tooo flakey anyway) > with a Xylogics Annex 4000. Then I can offer PPP...... but it's been > hinted that I need subnetting (or subnetting would make it easier!)
You don't generally *need* to subnet for PPP or the Annex, but it can be handy to have a bunch of networks. Our setup is 16/16 (0xff.ff.ff.f0) so we have one central network, and each of us has a network at home for the various machines resident there. > > Oh our network is not this neat & tidy thing!! We don't really > have many depts that are on their own (just the other bldg w/7 > machines (IP #s) & another group w/20 machines (IP #s). > and then there are thee PPP users .....
It depends a lot on your physical topology how you decide to subnet. What are you doing now - everybody on one network? a lot of unused addresses? Borrorwed IP number?
l & h, kev
>From digioia@sybase.com Mon Aug 7 01:24:53 1995 X-Sender: digioia@spanner To: Erin O'Neill <erin@skivs.ski.org> Subject: Re: subnetting a Class C In-Reply-To: <9508041956.AA23986@skivs.ski.org> Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Content-Length: 476 X-Lines: 12 Status: RO
Get O'Reily "managing TCP/IP" or "TCP System Admin."; I don't know exact name, as my copy is at the office right now, and I'm not.
-digioia
___________________________________________________________ David Di Gioia, sysadmin digioia@sybase.com Sybase, Inc. 415-943-8712 South Bay CS FAX: 415-943-8775 1945 Charleston Rd., Rm. 119 Mountain View, CA 94043, USA ___________________________________________________________
>From yli@libby.rbls.lib.il.us Mon Aug 7 08:26:28 1995 To: "Erin O'Neill" <erin@skivs.ski.org> Subject: Re: subnetting a Class C In-Reply-To: <9508041956.AA23986@skivs.ski.org> Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Content-Length: 2533 X-Lines: 57 Status: RO
On Fri, 4 Aug 1995, Erin O'Neill wrote:
> All I wanted to do was give my users PPP connections & I feel like > I've entered a dark maze! I'm trying to get the Powers > that Be to replace my ALM boards (which are much tooo flakey anyway) > with a Xylogics Annex 4000. Then I can offer PPP...... but it's been > hinted that I need subnetting (or subnetting would make it easier!) > > So I want to bite the bullet & subnet. I go to my O'reilly book > on Managing NFS & NIS. It's VERY cryptic (unusual for O'reilly). > So I go to my admin book (it's Sunos 4.1.2 & I use 4.1.3_u1). > It's not quite as cryptic as o'reilly but pretty close. Surely > there exists a paper that gives the nitty gritty details of > subnetting??? > > I also went to the summaries of this group & did a search on subnetting > & there wasn't anything there!! > H E L P!!!! > > Oh our network is not this neat & tidy thing!! We don't really > have many depts that are on their own (just the other bldg w/7 > machines (IP #s) & another group w/20 machines (IP #s). > and then there are thee PPP users ..... > > thanks ! > erin > =========================================================== > Smith-Kettlewell Eye Research Institute > Erin O'Neill email: erin@skivs.ski.org > 2232 Webster St. voice: (415)561-1612 > San Francisco, CA 94115 FAX: (415)561-1610 > =========================================================== >
Subnetting is a hard question. if you can give more specific infomation such as how many subnet do you need? what kind of network topology do you using? etc.
You also need to reconfig all the node on you network have a same subnet mask so they can talking to each ohter, that include your internet router. There are few books you can read, such as "Internetworking with TCP/IP volume1,2, and 3 by Douglas E. Comer" and "TCP/TP network administration by O'Reilly". Send me a e-mail, if you still need help.
\\|||// |^ ^| |0|0| ( _ ) /-----------------------------oOO--(-)--OOo---------------------------------\ | Yan Li Phone#: (309)799-3155 | l ASST. Manager, Automation Services Fax # : (309)799-7916 | | River Bend Library System E-mail: yli@libby.rbls.lib.il.us | | P.O.Box 125, 220 W. 23rd Ave. Coal Valley, IL 61240 | \--------------------------------OOO--OOO-----------------------------------/
>From stern@sunrise.East.Sun.COM Mon Aug 7 09:19:08 1995 To: erin@skivs.ski.org Subject: Re: subnetting a Class C X-Sun-Charset: US-ASCII Content-Length: 16080 X-Lines: 305 Status: RO
sorry you found the 1st edition of the book so cryptic; i'm rewriting that section as part of the 2nd edition. here's a draft of how it will look (taken from some columns from sunworld)
--hal .LP /sys.admin August 1992 .LP Broadcast News by Hal Stern .LP .nh If you're using a fairly small homogeneous Internet Protocol (IP)-based network, you may not have had to deal much with broadcast addresses and subnetwork masks. But as your installed base grows and you need more to add more networks, you'll find yourself tinkering with network masks and calculating broadcast addresses. In this column, we'll review how broadcast addresses get used, walk through some subnet mask and broadcast address calculations, and conclude by looking at some symptoms of misconfigured broadcast addresses and network masks. .SH Addressing Conventions .LP At the very bottom of the network protocol stack are the physical and data link layers: in most networks, these are represented in the real world by the Ethernet. Every machine on an Ethernet has a unique, 48-bit address known as an *Ethernet address* or a *MAC (Media Access Control) address*. Ethernet addresses are usually written as 6 colon-separated hex numbers, such as 8:0:20:a:6f:9. The first 24 bits of an Ethernet address are assigned to network hardware vendors, with the vendors assigning unique 24-bit suffixes to each piece of equipment that leaves their factories. The 24-bit prefixes are handed out in blocks by Xerox (creators of Ethernet). Every Sun, for example, has an Ethernet address that starts with 8:0:20, since Sun owns this prefix. .LP The special Ethernet address of all ones (ff:ff:ff:ff:ff:ff) is recognized by all machines - it is the Ethernet broadcast address. Normally, when one machine wants to talk to another, it fills in the Ethernet address of the destination in the packet's header. When a packet needs to be sent to all machines on the network, the sender fills in the Ethernet broadcast address, and all machines on the network choose to receive the packet, just as if it had been addressed specifically to them. .LP Networking and life would be remarkably simple if all machines existed in a single, flat network; however, physical limits in cabling and political constraints of our respective employers make this impractical. Instead of one large network, we have many physical networks, which are further divided into *logical* networks navigated by the Internetwork Protocol (IP). IP has its own addressing scheme: 32-bit numbers usually written as dot-separated decimal numbers, like 192.9.200.1. Every machine on a TCP/IP network has a unique Ethernet address and one or more IP addresses - one for each logical network to which the machine is connected. Note that IP addresses are *locally* assigned (each organization allocates them to its own machines), while Ethernet addresses are globally assigned (they are given out by the keepers of the Ethernet and the hardware vendors). .LP The 48-bit Ethernet address space is quite a bit larger than the 32-bit IP address space, and a single machine may have several IP addresses corresponding to a single Ethernet address. Enter the Address Resolution Protocol (ARP), a companion to the Internet Protocol used to map IP addresses into Ethernet addresses. Users and processes use the more convenient IP addresses, and let the lower layers of the protocol stack determine the Ethernet address for each destination. Examine your machine's IP and Ethernet addresses using *ifconfig*, and display the current mappings of IP addresses to Ethernet addresses using *arp*: .DS # ifconfig le0 le0: flags=63<UP,BROADCAST,NOTRAILERS,RUNNING> inet 138.31.32.26 netmask ffffff00 broadcast 138.31.32.0 ether 8:0:20:aa:45:67
# arp -a weekend (138.31.32.158) at 8:0:20:c:7b:2 whale (138.31.32.47) at 8:0:20:7:f4:d6 warbuck (138.31.32.15) at 8:0:20:2:d4:50 .DE *ifconfig* takes a network interface name as an argument. Determine the names of each interface on your system using *netstat -i*. .LP The Internet Protocol also supports the notion of a broadcast address. Unlike Ethernet, which has a single broadcast address, there are many IP broadcast addresses: one for each logical IP network. It's up to IP to recognize a broadcast address and convert it to the Ethernet broadcast address, instead of using ARP to locate a specific destination host. The Ethernet broadcast address of all ones is an addressing wildcard, since it matches all Ethernet addresses. The IP broadcast address for each network also uses wildcards to address all hosts on the IP network. The broadcast address is formed by taking a logical AND of a *network mask* and a machine's IP address. The part of the address that is masked out is filled in with a wildcard: either all zeros or all ones. In the example above, the network mask is shown as a 32-bit hex value, along with the result of masking that value against the machine's IP address to produce the broadcast address. .LP TCP/IP purists immediately point out that a *proper* IP broadcast address uses all ones to fill in the wildcard spot. Unfortunately, early versions of the Berkeley UNIX networking code did just the opposite, and an all-zeros based broadcast address. The convention was inherited by SunOS 3.x. To this day, Suns use a zeros-based broadcast address, like the one in the example above. The IP implementation in SunOS 4.x recognizes a zeros- or ones-filled wildcard as an IP broadcast address, but it still sets the masked out portion of the broadcast address to all zeros by default. Technically, a zeros-filled address is used to refer to a *network* itself, while the ones-filled address means "all hosts on the network". .LP Mixing the ones- and zeros-filled addresses may cause major interoperability problems if other hardware on your network requires one form or the other. Your best bet is to pick a convention - either zeros or ones based addresses - and make sure all of the machines on your network use it. We'll look at how to change from a zeros-based address to a ones-based address shortly. Before we change the broadcast address format, it's helpful to understand exactly where the network masks come from and how they are used to create subnetworks. .SH Network Masks .LP There are three default network masks, one for each *class* of IP address (Table 1). The "first octet" is the first number in your dot-separated IP address. Network masks are commonly shown as hexadecimal values, but you may also see them written "longhand" as dot-separate decimal values, just like IP addresses. For example, ffff0000 is sometimes specified as 255.255.0.0. The bits represented by a one in the network mask comprise the *network part* of an IP address while those that are masked by zeros form the *host part* of the address. The actual value of the network part is called the *network number*. In our first example, the network part is the first 24 bits of the address (represented by ones in the network mask); the host part is the last 8 bits of the address, and 138.31 is the network number. .LP The default network masks and default broadcast address may suit you quite well until you need another IP network. For example, you were using a Class C network (194.5.5.0, for example) and your network traffic grew to the point where you needed two separate IP networks. You *could* register for another Class C network number for the new network segment, or opt to break up the existing Class C network into multiple *subnetworks*. Creating subnets is simply a matter of creating a new network mask that has more bits than the default for the IP address class. .LP The Class C network mask has 24 bits of network number and 8 bits of host number. To create 4 subnetworks, you need 2 more bits of network number, which are "stolen" from the host part of the address. Making the network part 26 bits, instead of 24 bits, breaks the single 256-host network into 4 64-host subnetworks. So your new network mask for the Class C subnetwork is: .DS Hex Dotted ffffff00 255.255.255.0 original subnet mask 000000c0 0. 0. 0. 192 top two bits of host part --------- --------- ffffffc0 255.255.255.192 new subnet mask .DE Instead of a single network number (194.5.5.0), you now have 4 network numbers: 194.5.5.0, 194.5.5.64, 194.5.5.128, 194.5.5.192 (these values were formed by enumerating the possible values of the "new" network part bits). Each network has 64 IP addresses on it. .LP If you have a Class B address assigned to your site, you may choose to take 8 bits from the host part, making your network number 24 bits long. Since these 24 bit network numbers are also used by Class C networks, you usually hear this scheme called "subnetting to Class C addresses." Breaking up a Class B address into Class C subnetworks produces 256 subnetworks .LP Now that you have calculated a network mask, what do you do with it? The */etc/netmasks* file (and *netmasks* NIS map, built from this file on the NIS master server) is used to specify non-default subnetwork masks. The file contains network numbers and associated subnetwork masks (here shown in dotted form): .DS # Network netmask 192.5.5 255.255.255.192 138.31 255.255.255.0 .DE Note that the "keys" in the file are the original, unmodified network numbers - not the subnetwork numbers. Why? If the file was keyed on subnetwork number, you'd have to know the netmask *before* you even looked in the file. Since the purpose of the file is to specify non-default network masks, this is somewhat silly. Furthermore, all hosts on the network should use the same network mask, so specifying a single network mask for the entire "set" of subnetworks reduces the chances of a different network mask creeping onto some machines. .LP The *ifconfig* command executed out of */etc/rc.local* is where the network mask is applied to the IP address to set the broadcast address for the machine: .DS ifconfig -a netmask + broadcast + .DE The *-a* switch goes through all interfaces, configuring them one at a time. The plus signs (+) indicate that *ifconfig* should use the */etc/netmasks* file or the *netmasks* NIS map to look up the appropriate subnet mask for each interface. .LP The *ifconfig* utility forms the broadcast address by taking a logical AND of the network mask and the interface's IP address, and filling in either zeros or ones for the host part. In the introductory *ifconfig* example, the network mask 255.255.255.0 was applied to the IP address 138.31.32.25, producing a broadcast address of 138.31.32.0. If this machine used a ones-style host part, the broadcast address would have been 138.31.32.255. Mixing zeros- and ones-based broadcast addresses is usually a recipe for trouble, so it's best to make sure all machines on your network use the same form. .LP Switching To Ones .LP If you have other gear that requires an all-ones broadcast address, or if you want to test out your network using the "official" standard for IP broadcast addresses, you'll need to change the default addresses generated by SunOS. As an exercise, the conversion may prove to be more work than it's worth, but Solaris 2.0 switches to ones-based broadcast addresses, so the exercise may be a necessary step in preparing to perform a major operating system upgrade. .LP The simplest approach under SunOS 4.x is to replace the all-inclusive *ifconfig* in */etc/rc.local* with a set of *ifconfig* commands that explicitly list the broadcast addresses desired. For example, to set up network interface *le0* on network 138.31.32.0, and *le1* on network 138.31.50.0, use the following two *ifconfig* commands: .DS ifconfig le0 network + broadcast 138\\.31\\.32\\.255 ifconfig le1 network + broadcast 138\\.31\\.50\\.255 .DE [BILL: note that these should be single backslashes before each dot] The dots are escaped with backslashes because the */etc/rc.boot* script modifies the shell's input field separator (IFS) to include the dot as an argument separator. Without the backslashes, the broadcast addresses are separated into 4 values, instead of a single argument with embedded dots. .SH Cleaning Up .LP You've changed your network mask. You've set up ones-based broadcast addresses. Everything works (you hope). The most common problem with adopting a new network mask is that not all machines pick up the new mask. If a host with multiple interfaces doesn't have the correct subnet mask, it's likely that packets will get dumped onto the wrong interface. Consider a machine with IP address 138.31.32.26 on interface le0 and 138.31.50.26 on interface le1. If the subnetwork mask is set to 255.255.255.0, then this machine will have separate routes to the .32 and .50 subnetworks of Class B network 138.131. Use the default Class B network mask by mistake, though, and no traffic will ever go out over the le1 interface: the machine will see a single 138.131 network, attached to interface le0, and it will have no reason to use le1. .LP Check the routes adopted by machines after changing your subnet mask using *netstat*: .DS % netstat -r -n Routing tables Destination Gateway Flags Refcnt Use Interface 127.0.0.1 127.0.0.1 UH 2 698 lo0 138.31.32.0 138.31.32.26 U 0 483 le0 138.31.50.0 138.31.50.26 U 0 483 le1 .DE In this example, we've been subnetting the Class B address into Class C subnetworks. If this machine had an incorrect subnet mask, it would show the same destination attached to two (or more) interfaces: only the first configured interface would actually transmit packets. .LP Routing problems are notoriously hard to find, since they are often temporarily resolved by information from other routers and gateways on the network. If you experience intermittent failures, or periods when some machines just seem to vanish from the network, verify that all routers and gateway hosts are using the correct subnetwork masks. If you aren't using NIS to distribute the *netmasks* information, make sure that */etc/netmasks* is distributed regularly to all machines on the network. .LP The result of mixing ones and zeros based broadcast addresses can be much more spectacular: a broadcast storm. A machine that sends out a broadcast address not understood or accepted by other machines sometimes trigger a wave of broadcast requests and replies. Remember that it's up to the implementation of IP to decide if an IP address is a broadcast or a host address, and to only use ARP on host addresses. If a machine is configured with a broadcast address it can't recognize, it may send ARP requests for that broadcast address (thinking it refers to a host, and not an entire network). Broadcast storms last up to several minutes, and often bring network traffic to a complete halt like a multiple car pileup on the freeway. If you think you're suffering from broadcast storms, use *etherfind -broadcast* or a network analyzer to watch for broadcast packets, and verify that waves of broadcasts aren't triggered by a host using an unacceptable broadcast address. .LP It's heartening to know that UNIX features with apocryphal backgrounds, like zeros-based broadcast addresses, can be corrected over time (the 640k base memory limit in DOS may be around much longer). Solaris 2.0 uses ones-based broadcast addresses, so it may be worthwhile to convert your network over to the correct broadcast address form, and shake the bugs out before attempting a major operating system upgrade.
Table 1.
Address Class First Octet Network Mask A 1-126 ff000000 255.0.0.0 B 128-191 ffff0000 255.255.0.0 C 192-223 ffffff00 255.255.255.0
>From rali@hri.com Mon Aug 7 09:59:06 1995 Subject: Re: subnetting a Class C To: erin@skivs.ski.org Cc: rali@hri.com (Reto Lichtensteiger) In-Reply-To: <9508041956.AA23986@skivs.ski.org> from "Erin O'Neill" at Aug 4, 95 12:56:35 pm Reply-To: rali@meitca.com X-Mailer: ELM [version 2.4 PL23] Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 3040 X-Lines: 79 Status: RO
Back at the ranch, Erin O'Neill scribed:
<> So I want to bite the bullet & subnet. I go to my O'reilly book <> on Managing NFS & NIS. It's VERY cryptic (unusual for O'reilly). <> So I go to my admin book (it's Sunos 4.1.2 & I use 4.1.3_u1). <> It's not quite as cryptic as o'reilly but pretty close. Surely <> there exists a paper that gives the nitty gritty details of <> subnetting??? <> <> I also went to the summaries of this group & did a search on subnetting <> & there wasn't anything there!! <> H E L P!!!!
Erin,
Dunno if you've already gotten sufficient response to your "cri de couer"
Subnetting isn't that mysterious, though at first glance it does seem arcane ... Subnetting is simply using parts of the binary IP address that are, by default, part of the host value as a network value. (Did I say "simply"? <g>)
The netmask tells the system what part of the address to treat as the "network component" of the address and what part is the "host component".
When the system wishes to send a datagram (ie., a 'ping' ICMP Packet) it translates the hostname to an IP address and compares the network values. If the network is the same as it's own network it determines that the host is "local" (on the same piece of wire) otherwise it forwards the packet to the nearest router (usually set using a 'default route').
By 'stealing' bits from the host part of the address you can create "subnetworks" between the default network (Class C) and the hosts.
I'll use skivs.ski.org for the example:
dig tells me that skivs' IP address is 192.207.85.11 the default netmask for a class C is 255.255.255.0
In binary the address is: 110000001100111101010101 00001011 (eesh ...) and the default mask is 111111111111111111111111 00000000
| Here's the separator between host and network ---+
Now if you set a netmask of 255.255.255.192 it looks like this:
In binary the address is: 11000000110011110101010100 001011 and the default mask is 11111111111111111111111111 000000
| Here's the new separator between host and network -+
You will notice I used 2 bits of the last octet to create the subnet, this is because IP defines an "all zeros" address to be the network itself and "all ones" address to be the broadcast address. SO, we had to reserve some of what were formerly host addresses as "new" network reserved addresses.
That's really all there is to subnetting.
One catch that you must be aware of: If you create subnets, then your hosts need to be on the same subnet, or they'll need to have a router between them in order to communicate.
I hope this has been of assistance,
-Reto L.- -- R A Lichtensteiger rali@meitca.com -or- rali@hri.com http://www.hri.com/HRI/People/rali.html
System Administrator Mitsubishi Electric Information Technology Center Waltham MA 02154 (617) 466-8304
I use Solaris because someone told me it was admirable to work with the handicapped ...
>From leif@infoseek.com Mon Aug 7 12:52:14 1995 To: erin@skivs.ski.org (Erin O'Neill) Subject: Re: subnetting a Class C Content-Length: 1948 X-Lines: 51 Status: RO
>Erin O'Neill writes: >All I wanted to do was give my users PPP connections & I feel like >I've entered a dark maze! I'm trying to get the Powers
Been there, done that... It's a mess...
>So I want to bite the bullet & subnet. I go to my O'reilly book >on Managing NFS & NIS. It's VERY cryptic (unusual for O'reilly). >So I go to my admin book (it's Sunos 4.1.2 & I use 4.1.3_u1). >It's not quite as cryptic as o'reilly but pretty close. Surely >there exists a paper that gives the nitty gritty details of >subnetting???
Well, I don't know of any papers about subnetting, but I'm doing it on my C-networks. Basically I decided to use 6 bits for the host part, instead of the normal 8 bits. That "splits" a C-network into four sub networks, with 62 machines per network (you loose the broadcast, and the "network" address, as usual).
Six bits for the host gives you a netmask of 255.255.255.192.
So, a C-network of 204.162.96.0 would get these sub-netted networks:
204.162.96.0 (hosts are 1-62, broadcast is 204.162.96.63) 204.162.96.64 (hosts are 65-126, broadcast is 204.162.96.127) 204.162.96.128 (hosts are 129-190, broadcast is 204.162.96.191 204.162.96.192 (hosts are 193-254, broadcast is 204.162.96.255)
On my Sun's, this is pretty much handled automagically, once you configure the /etc/netmasks file. This is an example from my file:
204.162.97.0 255.255.255.192
If you look at your "ifconfig" statements in your rc file, they should have "broadcast +" and "netmask +". That tells it to calculate those parameters from the netmask statement in /etc/netmask.
IMPORTANT NOTE: Never have empty lines (anywhere!) in /etc/netmasks. I spend hours trying to figure out why all my ifconfig's where dumping core, only to find a trailing empty line in /etc/netmasks... :(
I hope this helps a little. If you have any specific questions, please feel free to ask. I can't guarantee I can answer them though! :)
Cheers!
-- Leif
>From Pschauss@aol.com Mon Aug 7 14:42:25 1995 To: erin@skivs.ski.org Subject: Re: subnetting a Class C Content-Length: 1698 X-Lines: 54 Status: RO
I set up my system (Solaris 2.4) like this:
In /etc/netmasks I have an entry:
199.233.107.0 255.255.255.224
Since 224 (the last segment in the mask ) is hex E0, (binary 1110000) this indicates that the upper three bits of the last address segment are interpreted as part of the network address. This allows me to have up to 7 subnets ( subnet 0 is reserved ).
Possible subnet addresses are
00100000 ( 32 ) 01000000 ( 64 ) 01100000 ( 96 ) ...
So my /etc/networks file looks like this
smithtown 199.233.107.32 okc 199.233.107.96 fr 199.233.107.64
Node addresses are created by adding the node address to the last segment of the network address. Thus node 199.233.107.33 is on the smithtown subnet and node 199.233.107.97 is on the okc subnet.
Each node will need information on how to reach nodes on the other subnets. On Solaris, this can be entered in the /etc/gateways file. On my network, the one on node 199.233.107.33 looks like this:
net 199.233.107.64 gateway 199.233.107.35 metric 1 passive net 199.233.107.96 gateway 199.233.107.35 metric 2 passive
These entries indicate that nodes on nets 199.233.107.64 and 199.233.107.96 are accessed through gateway node 199.233.107.35. The metric value of 1 indicates that 199.233.107.35 is directly connected to network 199.233.107.64. The metric value of 2 on the second line indicates that 199.233.107.35 can reach nodes on net 199.233.107.96 throuh another gateway.
You can also set up your routings with the route(1M) command. Its arguments are almost identical to the entries in /etc/gateways.
Hope this helps.
Peter Schauss pschauss@aol.com Gull Electronic Systems Div Parker Hannifin Corp Smithtown, NY
>From grevemes@VTC.TACOM.Army.Mil Mon Aug 7 15:04:09 1995 Subject: Re: subnetting a Class C To: erin@skivs.ski.org In-Reply-To: <9508041956.AA23986@skivs.ski.org> from "Erin O'Neill" at Aug 4, 95 12:56:35 pm X-Mailer: ELM [version 2.4 PL22] Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 2311 X-Lines: 55 Status: RO
The easiest way to give users PPP access it to assign each one an IP address. This eliminates the need for subnets, etc.
Option #2 is to assign an IP address to each individual modem port. This means you must do everything at a user level (i.e. POP mail, etc.) and treat each connection as a client and not a host. (Sendmail would have problems, etc. It would forward mail to the IP address as a HOST, but two different people could be that 1 IP address.) This doesn't affect connection-oriented things like X Windows which depend on the IP address but not the host information. (The specific IP address is not persistence between individual instances of connections.)
I have both types operating off this same machine at the same time. (Some PPP software needs a fixed address, some will allow negotiation.) I also have user who have hosts (unix boxes at home) and other where are clients (macs, windoze).
-seg
'Erin O'Neill' > > All I wanted to do was give my users PPP connections & I feel like > I've entered a dark maze! I'm trying to get the Powers > that Be to replace my ALM boards (which are much tooo flakey anyway) > with a Xylogics Annex 4000. Then I can offer PPP...... but it's been > hinted that I need subnetting (or subnetting would make it easier!) > > So I want to bite the bullet & subnet. I go to my O'reilly book > on Managing NFS & NIS. It's VERY cryptic (unusual for O'reilly). > So I go to my admin book (it's Sunos 4.1.2 & I use 4.1.3_u1). > It's not quite as cryptic as o'reilly but pretty close. Surely > there exists a paper that gives the nitty gritty details of > subnetting??? > > I also went to the summaries of this group & did a search on subnetting > & there wasn't anything there!! > H E L P!!!! > > Oh our network is not this neat & tidy thing!! We don't really > have many depts that are on their own (just the other bldg w/7 > machines (IP #s) & another group w/20 machines (IP #s). > and then there are thee PPP users ..... > > thanks ! > erin > =========================================================== > Smith-Kettlewell Eye Research Institute > Erin O'Neill email: erin@skivs.ski.org > 2232 Webster St. voice: (415)561-1612 > San Francisco, CA 94115 FAX: (415)561-1610 > =========================================================== >
>From vic@imatron.com Mon Aug 7 17:29:12 1995 To: erin@skivs.ski.org Subject: Re: subnetting a Class C Content-Length: 236 X-Lines: 13 Status: RO
Hi,
Nope sorry this is not the elusive answer....( sigh ) Just another ME TOO.....
So, if you do get some nice neat answer please let me know. or if enough intrest, please summerize and I will get the info there
-Thanx Vic
>From sdr@rdga3.att.com Mon Aug 7 21:56:10 1995 To: erin@skivs.ski.org Original-From: rdga3!sdr (S. D. Raffensberger 500622500 (RD)) Original-To: skivs.ski.org!erin Subject: Re: subnetting a Class C Content-Length: 143 X-Lines: 9 Status: RO
Erin
Try borrowing the book "TCP/IP Illustrated" by W. Richard Stevens (ISBN 0-201-63346-9).
It seems pretty complete.
Steve Raffensberger
>From kpi@hobbes.crc.com Tue Aug 8 09:11:43 1995 To: erin@skivs.ski.org (Erin O'Neill) Sender: kpi@hobbes.crc.com Subject: Re: subnetting a Class C X-Originating-Host: seth Reply-To: kpi@crc.com In-Reply-To: <9508041956.AA23986@skivs.ski.org> Mime-Version: 1.0 X-Mailer: BWMail for Windows Version 3.2 Content-Length: 1799 X-Lines: 44 Status: RO
In <9508041956.AA23986@skivs.ski.org>, Erin O'Neill wrote: >All I wanted to do was give my users PPP connections & I feel like >I've entered a dark maze! I'm trying to get the Powers >that Be to replace my ALM boards (which are much tooo flakey anyway) >with a Xylogics Annex 4000. Then I can offer PPP...... but it's been >hinted that I need subnetting (or subnetting would make it easier!) > >So I want to bite the bullet & subnet. I go to my O'reilly book >on Managing NFS & NIS. It's VERY cryptic (unusual for O'reilly). >So I go to my admin book (it's Sunos 4.1.2 & I use 4.1.3_u1). >It's not quite as cryptic as o'reilly but pretty close. Surely >there exists a paper that gives the nitty gritty details of >subnetting??? > >I also went to the summaries of this group & did a search on subnetting >& there wasn't anything there!! >H E L P!!!! > >Oh our network is not this neat & tidy thing!! We don't really >have many depts that are on their own (just the other bldg w/7 >machines (IP #s) & another group w/20 machines (IP #s). >and then there are thee PPP users ..... > >thanks ! >erin >=========================================================== > Smith-Kettlewell Eye Research Institute >Erin O'Neill email: erin@skivs.ski.org >2232 Webster St. voice: (415)561-1612 >San Francisco, CA 94115 FAX: (415)561-1610 >=========================================================== >
Let me suggest the 'other' O'Reilly book, TCP/IP Network Administration. It deals with this pretty well.
Kevin
-- Kevin Inscoe, Sys. Admin., Coleman Research Corp., Launch Systems Group Orlando, FL. kpi@crc.com - As a matter of fact it IS rocket science My views and opinions do not reflect the same for my employer or my country and definitely not the president! * * * GO ORLANDO MAGIC!!!
>From jwright@phy.ucsf.edu Wed Aug 9 16:02:57 1995 To: Erin O'Neill <erin@skivs.ski.org> Subject: Re: subnetting a Class C In-Reply-To: <9508041956.AA23986@skivs.ski.org> Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Content-Length: 614 X-Lines: 12 Status: RO
you don't *really* need to subnet just to add ppp lines. there are some nifty things that you can do if you subnet, but for 95% of the pc users they would never know the difference.
the annex manuals are fairly good about this. really all you need to do is assign an ip number to the annex, and an ip number to each of the serial ports. i've got it working pretty well here.
Jim Wright Keck Center for Integrative Neuroscience jwright@phy.ucsf.edu Box 0444, Room HSE-802 voice 415-502-4874 513 Parnassus Ave fax 415-502-4848 UCSF, San Francisco, CA 94143-0444
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:31 CDT