addressing scheme for lan segments

[bkoum]

hi

What the best practice for assing addresses between L3 devices connected all together like a star in a L3 switch. I noticed from some examples they use ip addressess with 32 subnet mask but i dont understand why. Is there any problem if i use ips from a network 24mask for example ?

thanks

[Giuseppe Larosa]

Hello Bill,

you need to use a subnet mask that is wide enough to accomodate all the devices connected to the subnet and to leave space for additional devices.

But as you note multiple solutions can be used if we are using private ip addressing. (RFC 1918)

A /24 can be used to accomodate 6 devices.

If we were speaking of public addresses the subnet should be the smallest possible because these are valuable resources.

Hope to help

Giuseppe

[Joseph W. Doherty]

Are these loopback addresses on the L3 device?

If so, and they are being used for management purposes, a /32 mask is known as a "host mask", and can be carried in a routing protocol like OSPF. Doing this provides more device IP addresses since you don't have to set aside some of the network address space for a network address or broadcast address. For example, a /30 uses a block of 4 IP addresses vs. /32 only using 1 IP address.

[Giuseppe Larosa]

Hello Joseph,

I hadn't understood the original poster mentioning /32 subnet masks.

This makes sense

Best Regards

Giuseppe

[bkoum]

hello

im sorry u have right... im trying to understand why some configuration examples use the /32 bit sm for connection between routers. Is this different from using a subneted network or is more practical ?

[Giuseppe Larosa]

Hello Bill,

usually these /32 addresses are associated to logical "always on" interfaces called loopbacks.

One /32 loopback can represent a node in a routing protocol.

Then some physical interface can borrow the loopback ip address using

int type x/y

ip unnumbered loop m

but in this case I would expect loop m to have a mask less then 32.

However, the more common use od these /32 ip addresses is as node identifiers.

May you provide an example of the usage of these /32 ip addresses so we can see if we have understood and answered to your question ?

Edit:

I mean usually these /32 node identifiers are reached using common subnets solving the routing problem.

Then once they are reachable

10.80.60.1 can be router1

10.80.60.2 can be router2

and so on

Hope to help

Giuseppe

[bkoum]

there is a router with 8 serial interfaces which connect 8 buildings . Im not sure what the best tactics for addressing these serials links with private ips.

one different subnet with /30 sm per single serial interface which gives me 2 host only?

one subnet with with /28 sm for covering 16 host ?

or these /32 sm ips from one network ?

thanks again

[Joseph W. Doherty]

For serial interfaces, you normally use /30 addresses with each device interface (on the two ends of the link) using the "middle" two addresses of the /30.

As Giuseppe describes, it's possible to define interfaces that borrow another interface's IP address, but with private IP addresses, there's often enough address space this isn't required.

If two routers are connected by Ethernet, you can also use /30, but if the Ethernet will be shared by other hosts, then increase the address space as needed. If you're working with 10.x.x.x private addresses, you might just use a /24 even if you only start with a low number of hosts.