jskfan
asked on
SUbnetting or Supernetting Strategy
http://jodies.de/ipcalc?host=192.168.0.0&mask1=24&mask2=
I found the Tool on the above link. It is helpful to calculate subnets and supernets, and it gives more details.
** If I understanding Subnetting a network gives better performance. However I would like to know the strategy of subnetting, I mean based on what objective should I subnet my Network, or just leave a large Network.
some Administrators they say Subnets should map physical location of resources and Vlans.
For instance if I have many locations , each location should have a separate subnet from other locations..That makes sense.
Let 's say I have 5000 users in 5 separate physical locations linked by WAN mediums...
300 users in location1, 1000 in location2, 700 in location3, 400 in location4, and 2600 users in location5(Corporate office)
At this point is the IP address class (A,B,C) is more important than the Mask for example (/19 )
I chose /19 because it gives 8190 hosts for whatever class I chose...
My point is how do I determine the number of subnets that will be in each location, and will satisfy the number of hosts (5000 users).
Any help will be very much appreciated
Thanks
I found the Tool on the above link. It is helpful to calculate subnets and supernets, and it gives more details.
** If I understanding Subnetting a network gives better performance. However I would like to know the strategy of subnetting, I mean based on what objective should I subnet my Network, or just leave a large Network.
some Administrators they say Subnets should map physical location of resources and Vlans.
For instance if I have many locations , each location should have a separate subnet from other locations..That makes sense.
Let 's say I have 5000 users in 5 separate physical locations linked by WAN mediums...
300 users in location1, 1000 in location2, 700 in location3, 400 in location4, and 2600 users in location5(Corporate office)
At this point is the IP address class (A,B,C) is more important than the Mask for example (/19 )
I chose /19 because it gives 8190 hosts for whatever class I chose...
My point is how do I determine the number of subnets that will be in each location, and will satisfy the number of hosts (5000 users).
Any help will be very much appreciated
Thanks
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
At this point is the IP address class (A,B,C) is more important than the Mask for example (/19 ) I chose /19 because it gives 8190 hosts for whatever class I chose...Class is only relevant if you're running a classfull routing protocol (RIP or IGRP). Otherwise, the only thing that matters is the mask.
Let 's say I have 5000 users in 5 separate physical locations linked by WAN mediums...In this example, you will most likely end up with at least 5 networks with the largest being 2,600 hosts (actually, at least 9 since the links will be networks as well). Because they are connected with WAN links, typically those will be connected with routers. While you could connect them as bridged links (L2TP, metro ethernet, etc.) this is not that common. One of the reasons is that broadcasts become of enterprise wide event. For example, a broadcast storm in location 1 is now a company-wide broadcast storm. But with routers connecting your locations, that broadcast storm is isolated to location 1.
300 users in location1, 1000 in location2, 700 in location3, 400 in location4, and 2600 users in location5(Corporate office)
My point is how do I determine the number of subnets that will be in each location, and will satisfy the number of hosts (5000 users).This is something we used to cover in design classes (not sure if it still is since I haven't taught the design classes in about 12 years). It used to be the size of our networks was limited by the size of the collision domain. In many networks, once you hit 100 hosts, performance suffered. Then switches eliminated collisions and we were able to create larger networks.
Then we ran into issues with excessive broadcast traffic. Since everyone on a network is also in the same broadcast domain, a broadcast sent by one host will be heard (and processed) by all other hosts. Some O/S's are a bit chatty with broadcast and will send one out every 60 seconds. If you have 5,000 hosts on a single network, that means every second, every host is receiving and processing 80 frames that it really doesn't care about. And those frames are also using bandwidth throughout the network.
Since routers do not forward broadcasts, we can reduce the size of the broadcast domain. But the cost is we end up with more, smaller networks.
The rule of thumb has been when broadcast traffic exceeds 20% of the total traffic, then the broadcast domain needs to be broken.
Will it work if it's 40% (or 50% or 75%)? Absolutely. But you will see a performance hit. Whether it's enough that it's an issue is something that you will have to decide.
As for terminology, supernetting is when we take multiple, contiguous networks and combine them to create a single network by moving the mask boundary to the left (using a shorter mask) and assigning those addresses (with the shorter mask) to the hosts.
When we do the same thing for the purposes of reducing the size of the routing table, it's referred to a "route summarization".
Forget about routing protocols, let's keep it simple, just for sake of debate I want to point one thing.
Even in static routing you can use supernetting for efficiency :
you can summary route
ip route 192.168.0.0 0.0.7.255 x.x.x.x
instead of use bunch of routes
ip route 192.168.0.0 0.0.0.255 x.x.x.x
ip route 192.168.1.0 0.0.0.255 x.x.x.x
ip route 192.168.2.0 0.0.0.255 x.x.x.x
ip route 192.168.3.0 0.0.0.255 x.x.x.x
ip route 192.168.4.0 0.0.0.255 x.x.x.x
ip route 192.168.5.0 0.0.0.255 x.x.x.x
ip route 192.168.6.0 0.0.0.255 x.x.x.x
ip route 192.168.7.0 0.0.0.255 x.x.x.x
That is also supernetting, isn't it?
Even in static routing you can use supernetting for efficiency :
you can summary route
ip route 192.168.0.0 0.0.7.255 x.x.x.x
instead of use bunch of routes
ip route 192.168.0.0 0.0.0.255 x.x.x.x
ip route 192.168.1.0 0.0.0.255 x.x.x.x
ip route 192.168.2.0 0.0.0.255 x.x.x.x
ip route 192.168.3.0 0.0.0.255 x.x.x.x
ip route 192.168.4.0 0.0.0.255 x.x.x.x
ip route 192.168.5.0 0.0.0.255 x.x.x.x
ip route 192.168.6.0 0.0.0.255 x.x.x.x
ip route 192.168.7.0 0.0.0.255 x.x.x.x
That is also supernetting, isn't it?
That is supernetting, isn't it?No, That's route summerization.
Exact same concept, just a different term for the application.
I was reffering to:
The process of forming a supernet is often called supernetting, prefix aggregation, route aggregation, or route summarization.
http://en.wikipedia.org/wiki/Supernetwork
The part that I forgot is:
Supernetting requires the use of routing protocols that support Classless Inter-Domain Routing (CIDR).
:)
My bad.
Route 192.168.0.0 0.0.7.255 is advertised to other routers, :) not pointing in that direction.
The process of forming a supernet is often called supernetting, prefix aggregation, route aggregation, or route summarization.
http://en.wikipedia.org/wiki/Supernetwork
The part that I forgot is:
Supernetting requires the use of routing protocols that support Classless Inter-Domain Routing (CIDR).
:)
My bad.
Route 192.168.0.0 0.0.7.255 is advertised to other routers, :) not pointing in that direction.
ASKER
** In the scenario I came up with, How would an Administrator shape up the Network for each location ?
Don mentioned that unless we use Classful Routing protocol such as RIP , the IP address class is not important but the Mask does..
Based on Don comments, I believe /19 is the mask that will be closer to 5000 users, it will gives 8190 Networks and Hosts, That might be the right number considering other devices that might need IP addresses....
First , Considering Nowadays Hardware (Routers, Switches, Printers, Servers, PCs, etc..) that have performant enough Network Adapters and Interfaces, do we need to make subnets of 100 or 200 devices on each subnet ? I know we cannot plan for performance by making small subnets only, but also by adding hardware resources( SSD,Memory,CPU,etc...)
What I need to know, is how to make enough subnets for each physical location based on the number of users I have previously mentioned ? and will meet the total of 5000 users for all locations ?
If I understand a Maximum host in a subnet is 254, so if I choose 192.168.0.0/19
that would give me 8190 Networks and Hosts.
First subnet 192.168.0.1 To 192.168.0.254
Second Subnet 192.168.1.1 To 192.168.1.254
Last Subnet : 192.168.31.1 To 192.168.31.254
Don mentioned that unless we use Classful Routing protocol such as RIP , the IP address class is not important but the Mask does..
Based on Don comments, I believe /19 is the mask that will be closer to 5000 users, it will gives 8190 Networks and Hosts, That might be the right number considering other devices that might need IP addresses....
First , Considering Nowadays Hardware (Routers, Switches, Printers, Servers, PCs, etc..) that have performant enough Network Adapters and Interfaces, do we need to make subnets of 100 or 200 devices on each subnet ? I know we cannot plan for performance by making small subnets only, but also by adding hardware resources( SSD,Memory,CPU,etc...)
What I need to know, is how to make enough subnets for each physical location based on the number of users I have previously mentioned ? and will meet the total of 5000 users for all locations ?
If I understand a Maximum host in a subnet is 254, so if I choose 192.168.0.0/19
that would give me 8190 Networks and Hosts.
First subnet 192.168.0.1 To 192.168.0.254
Second Subnet 192.168.1.1 To 192.168.1.254
Last Subnet : 192.168.31.1 To 192.168.31.254
ASKER
Then you divide each number of users on each location by 254 an this would determine how many subnet each location will need..... We may need one or 2 extra subnet on each location for other IP devices
ASKER
I could be wrong on my comment above, the Network might still have only one Broadcast address: 192.168.31.255 , if that's the case that will defeat the purpose of performance I guess.....
ASKER
I know I am going back and forth in figuring out the right way to do it...
Well I am thinking that in order to keep the broadcast to minimum (at least one broadcast by location) we can still can take 192.168.0.0 Network but with different Mask on each location...This will make a different Broadcast IP on each location....
But I am looking on how to make a separate broadcast for each segment( for instance for each 254 hosts maximum)
Well I am thinking that in order to keep the broadcast to minimum (at least one broadcast by location) we can still can take 192.168.0.0 Network but with different Mask on each location...This will make a different Broadcast IP on each location....
But I am looking on how to make a separate broadcast for each segment( for instance for each 254 hosts maximum)
There's not nearly enough information to answer this question. What kind of links will be between the locations? What kind of equipment will be used to connect the locations to the connections between the locations?
ASKER
WAN Connections? can be Frame Relay, MPLS, etc...
I am not sure what technology is used...
I am not sure what technology is used...
ASKER
The goal is to reduce Broadcast and increase performance.
ASKER
IN 192.168.0.0/19
that would give me 8190 Networks and Hosts.
First subnet 192.168.0.1 To 192.168.0.254
Second Subnet 192.168.1.1 To 192.168.1.254
Last Subnet : 192.168.31.1 To 192.168.31.254
I wonder if I can just create Vlans and this will take care of the Broadcast.
Vlan0
Vlan1
..
All the way to VLan31
that would give me 8190 Networks and Hosts.
First subnet 192.168.0.1 To 192.168.0.254
Second Subnet 192.168.1.1 To 192.168.1.254
Last Subnet : 192.168.31.1 To 192.168.31.254
I wonder if I can just create Vlans and this will take care of the Broadcast.
Vlan0
Vlan1
..
All the way to VLan31
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
Thank you All Guys!!
You are absolutely right in all what you argue but in the, let's call it, practical world there is little need to have routing protocols of the type OSPF, EIGRP and BGP represent for internal private IP infrastructures. It not only raises the requirements and complexity internally but also the costs regarding hardware, installation and maintenance over time. More complexity more to update and less hardware to choose from which generally means higher prices. I would argue that few Companies have the need for OSPF internally not to mention BGP in other than border routers/firewalls.
I would make an VPN Mesh infrastructure for all internal use with one or perhaps two entry/exits for redundancy then utilizing BGP and whatever other protocols necessary . Internally I would stick to those routing protocols that we find in some of the less expensive firewalls/routers given the fact that we, for most companies, have a fairly limited number of static networks and locations. This will have little impact on the overall performance but quite significantly address the costs! One base requirement is of course that EVERYTHING can be monitored and managed remotely and that "normal" practice regarding redundancy is applied!
Regards!