subnetting

[xtech]

Hi,

I am confused about subnetting - the 128 mask. I am using Boson guides, and they do not use a 128 on a class C. I thought the first address group would translate to this network and the last address group would translate to broadcast, so if you only have two groups you cannot use it on a class C. The cheat sheet also has the power of two examples, where usable address groups = address groups - 2, (first and last) We are on a 208.x.x.x/25, so I know it works.

I am also confused on the supernetting issue. Thanks - Wayne

[ggatten]

Remember that IPv4 addresses are 32 bits long - 4 octets, or 4 groups of 8 bits. If you break everything down in binary, it will make more sense.

In your example of 208.x.x.x/25 - that's a 25 bit mask or 255.255.255.128 in decimal; or

11111111.11111111.11111111.10000000 in binary - 25 bits = 25 places set to "1".

This yields two networks: 208.x.x.0 (hosts .1 - .126, BC=.127) and 208.x.x.128 (hosts .129 - .254, BC=.255)

Supernetting groups smaller networks into larger ones to limit routing table sizes. For instance your 208.x.x.x network:

Let's say your company has this entire address space - all 208.0.0.0/8 to itself. You then break that range up into 65,536 "class C" or /24 networks: 208.0.0.0 - 208.255.255.0. Now, if you own all these networks, and they are all only accessible through your kick-a$$ OC-192 Internet connection, would you advertise each /24 network seperately and take up 65,536 routing entries in thousands of routers around the world? No. In fact most ISP's would not allow you to do this.

Instead, you "SuperNet" everything and only advertise 208.0.0.0/8. Anything that starts with 208 goes to you. Now only 1 routing table entry is required. This also minimizes route flapping causing serious BGP stability issues - but thats another story.

It's tough enough to teach this face to face let alone this method. My best advice is look around - there's thousands of doc on the net about this - some better than others of course. Just get used to binary and it will make more sense. Also, grab a subnet calculater - like from WildPackets - and play with it; it may light the bulb for you.

HTH

Gary

[xtech]

Thanks

- Wayne

[Harold Ritter]

There used to be a restriction about subnet-zero (all subnet bits set to zero) and subnet all-ones (all subnet bits set to one).

These restrictions have been lifted since they don't really make sense in the classless context.

For more information please refer to the following document:

http://www.cisco.com/en/US/tech/tk648/tk361/technologies_tech_note09186a0080093f18.shtml

Hope this helps,