vlsm

hodgey87

hi all,

just wondered if anyone has any good information on vlsm

ive been using the todd lammles book but still cant really grasp it

thanks

:P

widjerd

variable length subnet masks -

Basically, in the past in a network people used one subnet mask for the entire network, for instance every single point-to-point link, host addresses and everything else would HAVE to have the sam subnet mask. IE, you could NOT have a network with one set of hosts using 255.255.255.0 and a different set of hosts using 255.255.255.192.

The disadvantage of networking like this meant that for every point to point link, which has to have its own network or subnet, would consume a huge amount of addresses that could then not be used. For example, if your network is using a /27 mask and you have 3 point to point links, you are having to use 3 different subnets for each point-to-point link. In each subnet for each point to point link you would be using 2 IP addresses but be wasting 28 addresses.

Therefore VLSM was introduced - It allowed you to conserve addresses by using different subnet masks in a network.

NOW, you can use a /30 address for point-to-point links therefore wasting NO addresses, and conserving a huge amount of addresses. This can also apply to host subnets, for instance if you have 12 hosts in one subnet and 30 in another, you can use a /28 for the 12 hosts (14 usable addresses, only wasting 2 addresses) and /27 for the 30 hosts (30 usable, no wastage of IP addresses).

RIPv1 and IGRP are classFULL routing protocols, and therefore do not route VLSM's.
Ripv2, OSPF and EIGRP are classLESS, and can route VLSM's.

Netstudent

While we're on the subject, I have a question.

Does the VLSM strategy call for starting with the smallest subnets and working your way up the blocks for larger subnets OR vice-versa?

I think I have heard of mixed strategies, but I wanted to get more input on someone's strategy they use to maintain a non-overlapping contiguous IP infrastructure.

lets say you are a Jr. Net Engineer and your in the design phase of a project and you know you need subnets that will provide an adequate number of hosts for:

LocationA-175hosts
LocationB-20 hosts
LocationC-65 hosts
LocationD-310 hosts
LocationE- 213 hosts

And then 4 point-to-point networks between routers, so /30.

Would I start with the /30 for the P-T-P or /23 for location D

widjerd

Netstudent wrote:

While we're on the subject, I have a question.

Does the VLSM strategy call for starting with the smallest subnets and working your way up the blocks for larger subnets OR vice-versa?

I think I have heard of mixed strategies, but I wanted to get more input on someone's strategy they use to maintain a non-overlapping contiguous IP infrastructure.

lets say you are a Jr. Net Engineer and your in the design phase of a project and you know you need subnets that will provide an adequate number of hosts for:

LocationA-175hosts
LocationB-20 hosts
LocationC-65 hosts
LocationD-310 hosts
LocationE- 213 hosts

And then 4 point-to-point networks between routers, so /30.

Would I start with the /30 for the P-T-P or /23 for location D

There are no set rules ( i think, but please correct me if i am wrong), but if you think about it, you really want to use the addressing space most efficiently. Starting with 4 PTP links would use from .0 up to .15, but for example you did not have a /28 address but only a /27 which jumps in 32's you would technically be wasting .16-.31 as you would have to start at .32.

rjbarlow

Absolutely I think You have first to obtain a /22 block of ip addresses, then when You are using routing protocols that support VLSM assign contiguous network numbers possibly all on single routers, surely never overlapping and then, finally, use route summarizzation on them.

EDIT /22 not /24.

Netstudent

I do see how starting with the smaller blocks would create greater flexibility as far as maintaining sound continuity and no overlaps.

So when one sits down to ponder a strategy, do you just have to hunt and peck untill you find a contiguous VLSM?

Also where did you get /22? that block is way to big for any subnets I have listed. /22 = 1022 hosts

I understand that you have to be very careful with the different multiples or the blocks.
So lets try this scenario shall we?

LocationA-175 hosts---->block of 256-2
LocationB-20 hosts

---

>block of 32-2
LocationC-65 hosts

---

>block of 128-2
LocationD-310 hosts---->block of 512-2
LocationE- 213 hosts---->block of 256-2

PTP-2 hosts--->block of 4-2
PTP-2hosts---->block of 4-2
" "
" "
Looking at this, it is impossible to use a class C...So lets use a Class B and start with the Smaller Blocks. LETS SAY WE WERE GIVEN:

172.31.0.0 255.255.248.0......Which gives us a possible 2046 IP's (172.31.0.1 - 172.31.7.254)

PTP1 = 172.31.0.1 - 172.31.0.2

---

> SUBNET 172.31.0.0 /30
PTP2 = 172.31.0.5 - 172.31.0.6

---

>SUBNET 172.31.0.4 /30
PTP3 = 172.31.0.9 - 172.31.0.10

---

>SUBNET 172.31.0.8 /30
PTP4 = 172.31.0.13 - 172.31.0.14

---

>SUBNET 172.31.0.12 /30

We can't use 172.31.0.16 subnet because the next block we need is a block of 32. Which would come out to be 172.31.0.16 /27 and with a /27 you get increments of 32 so that won't work since 16 falls in between RIGHT? SO WE SKIP IT? or does this prove we should have started with the larger blocks first?

LocationB = 172.31.0.33 - 172.31.0.62

---

>SUBNET 172.31.0.32 /27
LocationC = 172.31.0.129 - 172.31.0.254

---

>SUBNET 172.31.0.128 /25

So here again we have wasted 65 - 126 on the 172.31.0.64 subnet. IS THIS ACCEPTABLE?

Would anyone like to comment, am I right or wrong? Is this good or bad? Anyone want to continue? It's 5:30 time for me to go home.

=====================================
Back home

So now we have 2 locations that need blocks of 256 -2 or 254 however u want to look at it.

LocationA = 172.31.1.1 - 172.31.1.254

---

>SUBNET 172.31.1.0 /24
LocationE = 172.31.2.1 - 172.31.2.254

---

>SUBNET 172.31.2.0 /24

Now we need a block of 512-2 for the Last location. this comes to a /23...This gives us Increments of 2 (256-254).

Now the next subnet is 172.31.3.0, but we can't use it because a /23 is incremented by 2's. So we skip it, wasted it, and we'll use 172.31.4.0 /23

LocationD = 172.31.4.1 - 172.31.5.254

---

>SUBNET 172.31.4.0 /23


So we still have 172.31.6.1 - 172.31.7.254 for expansion because we were given 172.31.0.0 255.255.248.0

IS THIS RIGHT? If so I have to give myself a pat on the back because I have never actually worked through a long VLSM problem before.

rjbarlow

Netstudent wrote:

I do see how starting with the smaller blocks would create greater flexibility as far as maintaining sound continuity and no overlaps.

So when one sits down to ponder a strategy, do you just have to hunt and peck untill you find a contiguous VLSM?

Also where did you get /22? that block is way to big for any subnets I have listed. /22 = 1022 hosts
...
[CUT]

Well, I have fused at this hour, however You ad say that You needed to allocate 735 host, (i could suppose dislocated in some buildings...), then You have to obtain a /22 mask, because with a /23 You can allocate only 510 hosts maximum. Moreover with VLSM is not necessary to allocate all subnet as contiguous, but is better to do that on single routers for the the possibility to make route summarization on single routers. I could well fail, but that's miy think (now).
See You later.

Netstudent

Are you talking about the main large block that we would start with. If so, you can't just add up the IP's that are needed, you have to add up the blocks that would be required for each location.


This is a scenario where there is one Ethernet network on each router. So there would be no need for summarization unless there were many networks being advertised on each router right???

rjbarlow

Netstudent wrote:

Are you talking about the main large block that we would start with.

Yes.

Netstudent wrote:

This is a scenario where there is one Ethernet network on each router. So there would be no need for summarization unless there were many networks being advertised on each router right???

No, I thought to one scenario with more ethernet on single routers and discontiguous networks and one assigned IP subnet address from which starting, like that described into the chapter 7 of the ICND book of Odom (pag.236)
however is possible I mistook or that something escapes to me, but I have not never faced problems of net architecture and then, unfortunately, I can't say nothing more then already I made.
I think also the question You have asked goes beyond the CCNA (hope it).

Netstudent

Oh ok...Thats cool man, thanks for your input.

I'm still wondering if wasting 334 IP's out of the block I was (hypothetically) given 172.31.0.0 255.255.248.0 is acceptable.

In the real world, would you go back and re-work it?

I think thats a total of 1182 usable IP's allocated with 334 wasted.

rjbarlow

Netstudent wrote:

In the real world, would you go back and re-work it?

No, is not possible.

Netstudent

Well what I ment was, If this were an actual scenario in the real world, would you go back through the VLSM process to find a more effecient VLSM scheme.

I didn't mean go through the already configured network to reconfigure the whole network with different Ip's.

If we were still in the design phase, would this be acceptable?

rjbarlow

I can't now answer to you about that question, I'm concentrated on other parts of the CCNA curriculum like creating VLANS, then I don't want waste time on arguments that I think are not covered in the CCNA exam.