OSPF is a PITA. is-is isn’t.
OSPF requires operational and configuration overhead when you have 1000 layer 3 devices in a network, to maintain the configuration automation template etc, across IPv4 and IPv6.
Whereas with is-is, since it’s CLNP, I don’t need IP addressing whatsoever for underlay network, it can run unnumbered with a single-instance. Minimises operational and configuration expenses.
If OSPF is so great, why is there no Tier 1 carrier using it instead of is-is?
I do know that XDP/DPDK not only work for x86.Second hand E5 2699 V4 is only $136.54 in China, And mellanox 25G mcx4121a is only $20.50. If ROS support XDP/DPDK. I will built a X86 machine straightly and not consider any routerboard of fast path support. XDP/DPDK makes routerboard uncompetitive.
I did not say, that OSPF is in any way better than ISIS. It is still a fact, that in enterprise networks OSPF is still in use and scaling got much less an issue with as much resources we have today in any router.
So probably current support for routing protocols is no reason for MT to get used more in those networks. And Tier 1 Provider are really no market for MT.
In contrast: low IPv6 performance is very well a reason for not buying MT in SOHO and in Enterprise markets!
The majority of network engineers disagrees with you:
http://forum.mikrotik.com/t/is-is/27301/1
Hmmm, I did not read every post in that thread, but after a quick look:
THAT thread is about the benefits of ISIS and there the network engineers do discuss benefits of implementing ISIS.
I still never said anything about OSPF being better or ISIS being inferior.
In this thread the goal is to discuss IPv6 support and I believe IPv6 performance is BAD on MT. MT should try to do something about that, before implementing another routing protocol. Especially if they would speed up the whole data plane in the process!
That’s my opinion and a 100 network engineers could not beat it out of me.
At least I´d try to run away.
Old ER-4, and even older ER-Lite, were using Cavium SDK (proprietary) to offload established connections to hardware (Octeon SoC).
It isn’t pure SW optimization (like MT fast-track, or VPP), and it is not either a true L3HW. Something in between.
The complexity of this implementation was so high that EdgeOS v.2.x took years to reach stable, with a history of long-lasting bugs (packet re-ordering, ipsec, …), as SoC vendor (cavium) was not willing/able to provide modules for recent linux kernel.
Likely it was one of the reasons they terminated the EdgeRouter line.
I have to agree, as early as 2015 (even before ?), small ER-lite managed to forward 1Gbps (IPv4 + Ipv6) with near-zero CPU usage.
Of course this (the above) is also a lesson learned by MikroTik, and likely the reason why such things are not incorporated that quickly anymore.
See how it went with the long release delay of RouterOS v7 (updating the heavily patched Linux kernel was a major effort), and what is happening now with Wireless (maintaining the old wireless driver became too much work and now that a standard driver from the manufacturer is used, a lot of functionality is lost).
And MikroTik use many different CPU and SoC architectures and should they implement IPv6 fasttrack only for e.g. CCR20xx the howling here would be even worse than it is now.
+1 for IPv6 fasttrack and performance improvements! Thanks 
+1 for IPv6 Fasttrack
+1 !
+1, please
+1
Absolutely crucial feature!
We needed.
Thanks.
+100000000000
Please add it!
++++
It really takes too long…
+1. Janis alluded to IPv6 support being added over 8 years ago at MUM EU '16. Even for newer boards with beefier CPUs, knowing the router is doing more work than it has to and drawing more power for IPv6 is frustrating.
+1 for this.
The fact that even on a CCR2216, Mikrotik’s own numbers show it only capable of 3M PPS at a 64-byte packet size with 25 ipfilter rules (no fastpath, which is necessary for fasttrack) means that with IPv6 it’s a $2,500+ machine that can’t do full duplex 1Gbps line rate.
This is not acceptable in 2024 when nearly half the world is on IPv6 and there are entire countries that don’t have (or only have limited access to) IPv4.
Edit: I see folks saying by end of 2024 or in 2025 and want to say: Mikrotik has been talking about IPv6 performance improvements for close to a decade. I just wish they would be more honest about what’s stopping them from being able to implement it.
+1 for this~
IPv6 Fasttrack is coming once they work out platform issues … hopefully before end of this year or spring of 2025 …