routing benchmark

Radim Kolar kolar.radim at gmail.com
Sat Oct 6 21:09:58 CEST 2007


> > pomoci 10 ruli v linux fw se schova do chyby mereni. PF je znatelne
> > pomalejsi oproti IPFW.
> by mne zajimalo co s tim udela import pf as-of obsd4.2
s nejvetsi pravdepodobnosti s rychlosti vs Linux nic. Jeden fw jsem
pred par lety programoval a testovat adresy/porty u IP packetu je
trivialni. Pokud je to tak pomale, tak to nebude tim, ze testovaci
algoritmus je pomaly. Bude to s nejvetsi pravdepodobnosti context
switchema nebo zamkama.

http://www.openbsd.com/42.html
Huge performance improvements in the network stack, including:
    * In pf, store routing table ID, queue ID etc directly in the
packet header mbuf instead of using mbuf tags (which use malloc'd
memory). This yields a 100% improvement in pf performance.

Pokud je to pravda, tak by se pf timhle melo dostat tak na uroven
ipfw2. Na testovani nebudu mit v nejblizsi dobe cas.

mne taky zaujalo tohle:
OpenBGPD 4.2:
    * Include support for Four-octet AS Number Space.
    * Allow the use of pkill -HUP bgpd to reload the config.

openbsd ma kewl userland



More information about the Users-l mailing list