The discussion pops up again and again – someone says “but /64 subnets are such a waste! Surely we only need n bits?” (where “n” is any number smaller than 64). Then someone chimes in and says “we should have choice! Just remove that standard.” Someone else adds “you could do SLAAC in smaller subnets you know…” – and around we go again. This post is my take on the “/64 everywhere” idea, and let me say up front I am for it!
Those bemoaning the “waste” rarely seem to consider the advantages of the “/64 everywhere” approach. They decry the “waste” without looking at what that “waste” is buying:
- everything is the same. No more calculating (leaf) subnet sizes, no more robbing Peter to pay Paul, no more getting it wrong – every subnet is a /64.
- there are always enough addresses in any subnet for you add one more device. Or even a hundred. You might run into other limits, but the addressing will never get in your way.
- adding a subnet is easy – just take the next /64. No estimates, no calculations, no consideration or judgement needed.
- merging subnets is easy – you don’t have to make the receiving subnet bigger, it’s already big enough!
- your router configs are easier to understand
- your security configs are easier to understand
- your documentation is easier to write and easier to read
What this boils down to is greater ease of network management and administration. And that translates directly into money saved:
- saved in design time
- saved in implementation speed
- saved in ease of documentation
- saved in ease of troubleshooting
- saved in fewer errors made, corrections needed, or work redone
- saved in fewer security failures
- saved in less downtime
Spread across the whole Internet, all those advantages multiply.
All those benefits come just from having a standard subnet size, and one big enough to accommodate any likely number of hosts. The fact that it is big enough to handle completely unlikely numbers of hosts is a bonus – because you can forget about that problem too. Your subnets will always be “big enough”.
Perhaps we agree that a standard size network is a Good Thing, but disagree about how big it should be. After all, for any remotely reasonable number of devices in a /64 – 2, 200, 200,000 or even a million – it’s still zero percent to a dozen decimal places.
Let’s sliiiiiide things to the right, making the standard network smaller and smaller until that percentage is something bigger – say 1%. What have you achieved? If you still have a big enough subnet to accommodate any likely number of hosts, you are still “wasting” addresses. Anything smaller, and you’ve resurrected that old IPv4 concern “have I got enough addresses?” And if you make it so that you can have any sized leaf subnet, you lose all the advantages of “one size fits all”. And all this for what? To avoid “waste”.
My definition of “waste” is spending a limited resource for no benefit. The above benefits seem to me to have real value, so I don’t consider “/64 everywhere” to be a waste.