IPv6 Subnetting - Multi-Level Approach

If you have a large organization with many subnets, there may be more than one way you can divide up the /64 blocks you have available.

Say you have a /48 block, and operations in 13 cities. There are 7 departments in your company. The cities are:

0 - New York
1 - Chicago
2 - Los Angeles
3 - Paris
4 - Berlin
etc

The departments are:

0 - Marketing
1 - Sales
2 - R&D
3 - Administration
etc.

Assume each city has local versions of most or all departments.

First off, you can use hex digit based subnetting - no need for bit-based subnetting. You only need two levels. However, which splitting of the organization do you use for the top level subnetting (the one that determines the first hex digit of your Subnet ID), and which do you use for the second level subnetting (the one that determines the second hex digit of your Subnet ID).

Do you divide the subnets first by city, and then by department within each city? Or do you divide the subnets first by department, and then by city within department? On first look, it may not seem to make much difference.

In the first approach, the city determines the first hex digit of the Subnet ID, and the department determines the second hex digit. So, Paris Sales would use Subnet ID 3100, and Berlin Administration would use Subnet ID 4300.

In the second approach, the organizational unit determines the first hex digit of the Subnet ID, and the city determines the second digit. So Sales in Paris would use Subnet ID 1300 and Administration in Berlin would use Subnet ID 3400.

The answer is that it depends on your network architecture. Most companies would have a single route going into each city, and the departments would be split off within the city. In that case, it would lead to the best aggregation to use the first scheme (first level subnetting by city, second level by department).

If for some reason, there was a global network for Administration that covered all cities, and another for Sales, etc - then the second scheme would make more sense.

In general, arrange the levels in a multi-level subnetting scheme so as to aggregate network routes as much as possible. Generally the highest level geographical division (e.g. country) should determine the first hex digit of the Subnet ID, and smaller and smaller divisions should determine successive hex digits. But to be correct, you need to understand your organization's network architecture.

For a four-level scheme, the top level might be country, the second level city within country, the third level building within city, and the fourth level, floor within building. You could use hex digit based subnetting if each level had 16 or less divisions.

However, you might need to accommodate 12 countries, up to 7 cities within each country, up to 6 building in each city, and up to 50 floors in each building. In that case, you could use the following bit based subnetting:

4 bits for country (up to 16)
3 bits for city within country (up to 8)
3 bits for building within city (up to 8)
6 bits for floor within building (up to 64)

That is a total of 16 bits. In reality, you probably would use a single /64 subnet for a given building. So, you could use a three level scheme as follows:

7 bits for country (up to 128)
4 bits for city within country (up to 16)
5 bits for building within city (up to 32)

Again, a total of 16 bits. This would leave a lot more room for growth over the next decade, without having to redesign and renumber your network. Always try to think ahead to avoid having to do that, if at all possible.