Forwarding Protocols • Re: IS-IS

I'm a bit mixed on the new high core count moves. Frankly, for pure routing the 2004 and 2116 are blazing fast, optimizing routeros to use more cores would just keep making those better and better.

The biggest problem I'm seeing still comes down to single core performance. BGP is way faster, but still can get stomped handily by a FRR box on a modern intel chip. A shaping tree get's stuck in a single CPU core. SNMP gets stuck on a single CPU core.

Granted, you could work through some of those limitations by making them multi-threaded but some things are just not well suited for it. A shaper tree for instance sort of needs to have the top level HTB on a single core or some careful balancing of average traffic which would need a separate and new tool from mikrotik, and BGP needs to have each peer on a single core because moving data between cores is so much slower as to negate multi-threading gains.

I think it's a long shot but mikrotik dropping a 13-14th gen xeon box or similar ryzen box with really high core speeds would be awesome. I know there are ways to get routeros on that hardware, CHR, hacking x86 installer into a CHR on bare metal etc, but I'm not really interested in that. I want a *MIKROTIK* box in 1U with port options etc. Like a CCR2004-* but with a very high clock high performance CPU. Frankly, the only ARM series chip that would be interesting here is Apple's, everyone else is well behind on single core performance. Frankly, the i3-13100F would be so incredible in a ccr2004-12SFP+ format...

Statistics: Posted by PortalNET — Wed Jan 03, 2024 4:21 am