For years, we have heard of the depletion of the pool of available IPv4 address space on the Internet. At the same time, more and more networks are coming online. Out of these new networks, we have an ever-increasing number wanting to participate in BGP routing. This can be for several reasons.
The first is for what BGP was designed for — autonomy. Companies are not at the mercy of a single provider for critical infrastructure. If you have ever had to switch providers with IP addresses from the outgoing ISP, you know how much of a pain this can be.
So, what exactly is the problem? To participate in BGP, the accepted norm is to have an IP block of a /24 (256 IPs) or more. There is really no technical reason why this has to be this way. When router memory was expensive, a /24 was the cut-off for the smallest IPv4 block size you would allow through. This was to save memory and establish Mutually Agreed Upon Norms for Routing Security (MANRS). Most companies that are good at BGP have filters designed to drop prefixes smaller than /24s. This has been the accepted norm forever.
A company has two ways of getting IP space. As mentioned above, the first is to go to your ISP and have them rent or assign you IPs. The other is to go to a Regional Internet Registry and request an IP block. Either way, you must get a /24 to participate in BGP. Whether you use two or 200 IPs, doing BGP must still be a minimum of a /24.
So why is this important? Many companies wanting to do BGP don’t want to waste a /24 for a handful of IPs. Smaller Wireless Internet Service Providers (WISPS) only need a handful of IPs for NATting customers on the IPv4 front, especially if they are doing IPv6.
I propose now is the time to change this. The only reason you can not accept anything smaller is due to the agreed-upon norms. Routers have become more powerful and are shipping with more RAM. RAM is a commodity. The manufacturers should stop making devices with minimal amounts of TCAM memory. It’s cheap compared to 15 years ago. If you have a struggling router with low memory, maybe it is time to upgrade or change your design.
The rest is just getting the community to accept, say, a /25 or even a /26 as the smallest block. Yes, this means more work. Yes, it means the global routing table will get bigger. There are many ways to implement a change to the norm. A simple cut and replace for ‘le 24’ to ‘le 25’ would go a long way.
Justin Wilson is an ISP veteran, with over 20 years in the field. He’s a regular panellist and speaker at several Wireless Internet Service Providers Association (WISP) and Mikrotik conferences, a founding member of MidWest-IX based in Indianapolis, and the CEO of MTIN.NET LLC, a consulting firm specializing in ISP and enterprise markets.
Adapted from the original at Packets Down Range.
The views expressed by the authors of this blog are their own and do not necessarily reflect the views of APNIC. Please note a Code of Conduct applies to this blog.
Great write up!
Very logical
Yes, simple and useful. Why not!
The community needs to make this change sooner the better.
What you are really asking for is for the Internet as a whole to accept your sub /24 announcement as BGP doesn’t care and possibly your upstream provider doesn’t either. The case where it is important is when you are using your upstream provider’s space and want to announce it to someone else. In that case how do you prove you have the right to use it and what happens if you want to no longer be a customer of the first ISP?
As soon as you change those norms, the route table will grow exponentially… Believe me, this is not something that can be solved only with RAM.
Years ago, someone was a candidate for Ripe. What it promised was IPv5. it looks like this…
I agree 1000%! This was the reason I’m still postponing my own AS. I could very well work with /28.. the problem now is that /24 address space itself is more expensive than the peering port costs! that’s ridiculous