Date: Tue, 20 Mar 2001 10:53:23 -0800 From: Rich Morin <rdm@cfcl.com> To: freebsd-stable@FreeBSD.ORG Subject: Re: NFS performance Message-ID: <p05001905b6dd5542a7e2@[192.168.168.205]> In-Reply-To: <Pine.BSF.4.33.0103201041190.3344-100000@sdmail0.sd.bmarts.com> References: <Pine.BSF.4.33.0103201041190.3344-100000@sdmail0.sd.bmarts.com>
next in thread | previous in thread | raw e-mail | index | archive | help
At 10:45 AM -0800 3/20/01, Gordon Tetlow wrote: >Why are you using TCP? If you are on a reliable LAN, use UDP. TCP should >be used for long haul NFS. There are lots of reasons for using UDP, if you >want me to go into them, I will. Although TCP imposes some overhead, it may provide better worst-case performance than UDP. Several years ago, I chatted with a friend (Stan Hanks, IIRC) who had been trying to figure out why X terminals were doing better than diskless workstations. It turned out that, when the Ethernet started to get overloaded, some packets would get lost. The (UDP-based) NFS code would then attempt to re-transmit the entire 8 KB block. Given that the net was already overloaded, this would typically fail. The X terminals, meanwhile, were able to get their packets through. It seems to me that TCP-based NFS would fare better in this scenario, because it would only retransmit Ethernet packets. I'm not sure whether the moral of this story is to use NFS, however; a more reasonable strategy might be to ensure that the net never gets that heavily loaded! -r -- http://www.cfcl.com/rdm - home page, resume, etc. http://www.cfcl.com/Meta/md_fb.html - The FreeBSD Browser email: rdm@cfcl.com; phone: +1 650-873-7841 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?p05001905b6dd5542a7e2>