IPv6 is insane!

RouterOS General
1

While looking at some public routing information about a question someone else asked here, I found a company called Sucuri who has 7 /24 prefixes in IPv4…

But they have a freaking /29 in IPv6. Have the folks at RIPE lost their minds?

Sucuri has only announced seven /48 prefixes into IPv6 routing - obviously there’s one for each /24 of IPv4…
Dear God, that’s wasteful. (0.001% of their total space is even being announced on the Internet, which doesn’t even address how much unused space exists in each /48) I’m not blaming Sucuri (well, I am somewhat) because hey, they applied for some space and they got it… but is this kind of exorbitant allocation policy going to burn us later?

To put this in perspective, a /48 gives 65,536 networks to set up… okay, that’s cool - there sure is lots of room in IPv6… but they were given a /29… a /32 has as many network prefixes as there are IPv4 addresses - all the way from 0.0.0.0 to 255.255.255.255 - and Sucuri has 8 times more prefixes than a /32.

They have 34.3 b-b-b-BILLION prefixes to play with. (34,359,738,368 prefixes to be precise)

They’re only announcing ranges that account for 458,752 prefixes.
This means that even if they pack all seven of their /48 prefixes to 100% load, they’re only utilizing 0.001% of their space.

To make matters worse - Sucuri has chosen to use blocks that are arranged in a way that their allocation couldn’t be cut down easily to a /40 or even a /36 - no, they’ve got networks that vary at the 35th bit, so a /35 is the longest their prefix could ever be shortened to without them having to re-number two of their /48s.

Am I the only one who finds this sort of thing to be grossly wasteful?

I mean, I know there’re more IPv6 addresses (there are vastly far fewer networks in this space) than there are stars in the universe, or even particles in the universe (I think) - but with THIS kind of wasteful allocation behavior on the part of the RIRs, it looks like we could be burning through them a lot sooner than we should.

It’s like Bill Gates spending ten million dollars on a Big Mac or something… Sure he can afford to do that for a little while, but he’d blow all his money very quickly if that’s what he always did.

2

All LIRs get a /29 from RIPE, does not matter if you are an individual person (like me, have my own /29 too) or a big company (they can get even more space). Minimal allocation size at RIPE currently is /32.

3

ARIN’s base initial allocation for ISPs is /32, and of course you can get more than that with justification. It’s obvious that what RIPE’s doing behind the curtain is to silently reserve the adjacent /29 as “expansion” so that if you sufficiently utilize the /29, they’ll just bump your prefix up to a /28 and have done. They’re trying to keep things aggregated.

Even the mighty Google appears to have about 26 /32s worth of space… at least based on their BGP advertisements, anyway… only 3 times more than a /29 lol.


I’m going to go read the rules on what it takes to be considered an LIR by RIPE - my curiosity is piqued…

4

Well, I didn’t get very far digging into what it takes to be an LIR but apparently not a whole lot is required…

But I see what their reasoning is for the /29 allocations. The LIR concept is part of the hierarchical plan for IPv6 allocation in order to keep addresses as aggregated as possible.

I’m not so convinced about the feasibility of this concept because the goal is to minimize the number of prefixes in the global IPv6 routing table, but I think the way that the network is built and is increasingly non-hierarchical in topology is going to make this “keep things aggregated” policy into a pipe dream.

In fact, my example in the original post is being a poor steward of their LIR position - they’re announcing individual /48 prefixes all willy-nilly, when the whole reason they’ve been given a /29 is because in theory, they should announce this /29 and nothing else, and anyone they assign addresses to should be somewhere inside their network topology so that all assignments stay aggregated.

I think that as hardware continues to improve its capability, a global routing table of a few million routes is not going to be as daunting as it would have seemed back in the 90s when this stuff first started becoming standards. CIDR was too taxing for routing hardware when IPv4 hit the scene, thus the class A/B/C system, which was thrown out because it was too limiting. I suspect that the LIR concept is doomed from the start. Only time will tell.

5

sounds like some kinda of IP(v6)resources “cyber-squatting” or corporate/government of similar or worser intent.
personally i wouldn’t blame RIPE for that, they not really that big to see anything in everywhere, but reporting and having Reasonable explanation - would make sense.