From: Scott Michel <scottm@aero.org> To: Hugh LaMaster <lamaster@nren.nasa.gov> Cc: Hirofumi ABE <abe.hirofumi@lab.ntt.co.jp>, Kenjiro Cho <kjc@csl.sony.co.jp>, freebsd-atm@freebsd.org Subject: Re: Traffic shaping on HARP Message-ID: <38BD67A8.B805C672@aero.org> References: <Pine.GSO.4.05.10002291756110.4297-100000@kinkajou.arc.nasa.gov>
next in thread | previous in thread | raw e-mail | index | archive | help
Hugh LaMaster wrote: > Cisco ATM MTU is 4470 by default; if you want 9180 you have > to state it explicitly. I believe that setting it to 9180 > can exacerbate problems with fast SRAM buffer exhaustion on > some configurations with lots of subinterfaces. POS also > defaults to 4470, though I believe that it can be set higher, > also to 9180, and the Cisco GigabitEthernet max MTU on some > new interfaces is 4470 (1500 on others), so, it seems that 4470 > is kind of a Cisco "standard". Do you need 9180 for a particular > reason or would 4470 do? (I think 9180 might be the Fore ATM > default? Does 9180 show up in an RFC somewhere?) 9180 is kinda weird anyway when dealing with Sparc/Solaris boxes, since it's an odd multiple of the h/w page size. Causes strange effects in the Solaris kernel memory manager, at least on 2.5 and 2.6. Not that 4470 is much better either for the x86 or Sparc. So setting MUT to something that resembles your platform's page size is something I highly recommend. Yes, it's extra twiddling, but hey, this is ATM we're talking about -- you're already taking a network mgmt hit to maintain two infrastructures. -scooter To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-atm" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?38BD67A8.B805C672>