And isn't it true that the speed of a single core is very important for routing daemons (control plane), but less so for packet forwarding (data plane), which can be parallelized between cores, especially in Linux?
niedz., 2 lis 2025 o 07:05 Mike Neo <[email protected]> napisał(a): > As far as I understand it, BIRD 3 implements multithreading by allowing > different protocol instances or routing tables to run on separate worker > threads, but each individual bgp protocol instance itself still operates > mostly in a single-threaded manner. Right? > > niedz., 2 lis 2025 o 00:24 Alarig Le Lay via Bird-users < > [email protected]> napisał(a): > >> Hello, >> >> On Sat 01 Nov 2025 21:48:22 GMT, Mike Neo wrote: >> > Hi, >> > >> > What's more important for a bird, the clock speed of a single processor >> > core or the number of physical cores? >> > For example, which is the better choice: >> > 8x2.2 GHz or 4x3.8 GHz >> > >> > Kind regards, >> > Mike >> >> I always prefer the performance per core. Even if bird3 is >> multi-threaded, you’ll always have some locks between the threads (eg. >> compute and actual next-hop from BGP using OSPF). >> And on pure routing, the higher the frequency is, the fastest the packet >> is routed. >> >> -- >> Alarig >> >
