blackhawk

I having trouble getting my head around this.

Diagram shows four subnets (one at each corner, connecting to a box in the middle, could be router, doesnt say)

Class C Network 192.168.0.0 (IP`s just for purpose of question.)

Default subnet 255.255.255.0 the first 3 octets (24bit) are the network and the last octet (8bits) are the host, wich gives me 254 hosts or addressed space, I wont to brake up the 254 hosts into four departments (4 Subnets), so I change the last octet to .192

Mum.
192.168.101.1
255.255.255.192

Dad.
192.168.101.2
255.255.255.192

Son.
192.168.101.3
255.255.255.192

Daughter.
192.168.101.4
255.255.255.192

My questions,

Why are the subnet numbers 255.255.255(.192) kept the same for the four subnets ?

Why isnt the last octet (.192) different (190, 191, 192, 193) for each subnet ?
I understand that the more bits you use, the fewer hosts per subnet.

heatwave

hmm, thats a really bad example. A subnet mask just shows you what IP addresses are in your subnet. The problem is your not thinking in binary. Try this tutorial

vega55

192 = 11000000 in binary

So you will be using the last 6 bits as host addresses.

That subnet mask allows for (2^6)-2 hosts per subnet, and (2^2)-2 subnets of the class C network. So that subnet mask won't let you get enough subnets for what you want. Remember that you need to allow for the broadcast subnet and zero subnet. (I am coming from a Cisco CCNA perspective here anyway, and Cisco recommends as a best practice to not use those subnets).

Assuming that you do use the zero and broadcast subnets it would look something like this:
2^6 = 64 - so we need to count by 64 to get the subnet addresses

192.168.101.0
192.168.101.64
192.168.101.128
192.168.101.192

So to assign the first available address from each subnet to a different department you would need to do:
192.168.101.1
192.168.101.65
192.168.101.129
192.168.101.193

When you do a binary AND with any of those subnet address and the subnet mask you should end up back the the subnet address:
192.168.101.193 - 11000001 (just the .193 in binary)
255.255.255.192 - 11000000 (just the .192 in binary)
11000000 (.192)

192.168.101.129 - 10000001 (.129)
255.255.255.192 - 11000000 (.192)
10000000 (.128)

And so on.

The subnet mask can only ever be a power of 2. So you can use .128 .192. .224 .240 .248 .252 .254 .255 as valid subnet masks. Anything else is impossible because it needs to borrow bits starting from left to right, and can't skip bits.

heatwave

You can drop the -2 on the subnets this was only used for legacy IP addressing support (compatibility with legacy non-cisco devices). In cisco there is a command called "IP subnet zero" which is enabled by default that allows for the use of the all 1's and 0's subnets and is used heavily in VLSM/CIDR. In CCNAv1 test you had to subtract the -2, in CCNAv2 test you would get the answer wrong if you did -2, I never took the CCNAv3 test, but can tell you the CCNP tests will mark you wrong if you use the -2.

quote from cisco...

vega55

It is some tricky business deciding on whether to drop those two subnets or not. I just recently got my CCNA (last May), and everything I read in preperation said to drop the two subnets. In the accompanying practice tests you needed to drop them too. I don't recall having any number of possible subnets on my actual exam so I couldn't say how it would have actually works out.

In real life, as I mentioned and worked out they are available, and in fact are actually available by defauly on Cisco equuipment. So why they tell you to study for some exams without them, and others with I will never know. My guess is that as network addresses grow shorter they will change over to use of all available subnets and stop the nonsense of drop two.

blackhawk

Now Im really curious what materials you used. I just pulled out my CCIE book from cisco press "routing TCP/IP Volume I" written in 1998 and shows to start using the all 1's and zero's networks as VLSM has made it possible to distinguish between the networks. Page 288. as well in the CIDR section it says not using the all 1's and zeros for supernetting could cause routing loops.

heatwave

Out of curiosity what study guides did you use for your ccna that still says to do this? the modification to this is in RFC 1878 which was written in 1995 and quoted in the RFC is

"Many vendors only support subnetting based upon this
premise. This practice is obsolete! Modern software will be able to
utilize all definable networks."

keep in mind that was written in 1995 so modern software means after 1995 LOL!

There should be no confusion on when to not use these subnets basicly only if you have legacy pre 1995 software thats about it. It is note worthy to mention this for legacy applications but the use of VLSM / classless routing (also written in 1995) assumes the use of the all 1's and 0's subnets as the whole point was to expand the life of IPv4 addressing (why waiste 2 subnets?). basicly what Im trying to say is using ip subnet zero is part of using VLSM/CIDR. they go hand in hand.

vega55

I honestly can't really remember where I read it. My main source of study information was the Cisco Press CCNA Study Guide. But I also took some of the CCNA Networking Academy courses while I was still in school (only a year or two ago) too. It might have been there that I read to not include those subnets, but they do exist and are usable. Also I do think that the requirments for the CCNA, CCNP, and CCIE are slightly different so that might be a cause too.

blackhawk

Jiggy my lad, did you get this worked out yet??

You smart guys sure like to make simple things difficult

A subnet mask simply defines how many IP licenses you can have! A class C license (255.255.255.0 ) allows you 256 total licenses with two being used as described above.
A class B license (255.255.0.0 ) allows about 65000 licenses (254X256). A class A license (255.0.0.0 ) would be 254X256X256.
You do not change the segment values of a subnet mask to gain different sub-networks.

heatwave

simple answer is 255.255.255.0 is 11111111.11111111.11111111.00000000 in binary you barrow two bits from the host portion qand get 1111111.1111111.1111111.11000000 or 255.255.255.192 the last 2 bits you barrowed can be toggled on or off in combinations to create the subnets, the mask tells the computer how far in the address to look for a network address.
binary compilation = one octet = eight 1's or zero's (a bit) mathmaticly each bit represents a decimal number so starting from LEFT to RIGHT each bit equals
128, 64, 32, 16, 8, 4, 2, 1

the first two bits added together in decimal is 128+64=192 which is your new mask. so 192 says look at the first two bits to determine network portion, we do this conversion because computers only know binary and humans like decimals so adding human decimals together tells the computer which binary bits to use. so a mask of 224 would say look at the first 3 bits in any combination.(left to right) ie 128+64+32=240

this is why the mask stays the same since 128+64 will always be 192

Im not trying to make things difficult it just struck me odd to see someone still subtracting the zero and 1's subnets which will later confuse people. Im not trying to be nit picky either I think someone new to networking would appreciate the correct terms and use of current IPv4 addressing. Using the terminology of classes A,B and C ranges is also legacy as described by bassman. This was also eliminated by CIDR/VLSM which is now classless routing. you could not call up and request a class C range of addresses any more the correct current terminology is to use slash definitions. for example 192.168.0.0 mask 255.255.255.0 would be written 192.168.0.0/24 pronounced "192.168.0.0 slash 24" so basicly there really is no subnetting of classful addressing it is rather a classless IP addressing scheme. (CIDR = classless inter-domain routing) Looking at it this way I believe eliminates confusion for new networking people as thinking they have to subnet a class range when in reality the internet is classless, so when you call up for a /24 address space it very well could be 12.1.1.0/24 that you recieve.

quote from RIPE


"Can I obtain a class C address block?

Classful addressing (class A, class B, class C, etc) is now redundant and the RIPE NCC no longer uses that terminology. The original model for distributing IP addresses, based on classful addressing strategies, did not take into account the massive expansion of Internet use and was unable to scale to meet this expansion. Classless Inter Domain Routing (CIDR) is now one of the fundamental requirements for eligibility to receive IP address ranges."

Dont get me wrong though I do beleive the classful system should still be taught for legacy applications like RIPv1, but should be taught as such legacy not as a current design strategy. as It is not uncommon today to walk into a site that uses 10.0.0.0/24 as their addressing scheme.