From owner-freebsd-stable Mon Jul 30 7:33:53 2001 Delivered-To: freebsd-stable@freebsd.org Received: from cs.rpi.edu (mumble.cs.rpi.edu [128.213.8.16]) by hub.freebsd.org (Postfix) with ESMTP id EB17837B403; Mon, 30 Jul 2001 07:33:46 -0700 (PDT) (envelope-from crossd@cs.rpi.edu) Received: from cs.rpi.edu (monica.cs.rpi.edu [128.213.7.2]) by cs.rpi.edu (8.9.3/8.9.3) with ESMTP id KAA34464; Mon, 30 Jul 2001 10:33:44 -0400 (EDT) Message-Id: <200107301433.KAA34464@cs.rpi.edu> To: Cy Schubert - ITSD Open Systems Group Cc: "Matthew E. Martini" , Bill Paul , stable@FreeBSD.org, developers@FreeBSD.org Subject: Re: ypserv In-Reply-To: Message from Cy Schubert - ITSD Open Systems Group of "Mon, 30 Jul 2001 05:46:49 PDT." <200107301247.f6UCl1L17354@cwsys.cwsent.com> Date: Mon, 30 Jul 2001 10:33:43 -0400 From: "David E. Cross" Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG The problem with ypserv is definitely exacerbated with high load (I think it is a problem in the buffer management code in berkley db) a home network simply cannot generate the load to cause this (unless your home net has about ~40 machines on a switched segment, and the machine connected to 3 such segments over different interfaces ;) I also will point out a warning on www.sleepycat.com about using Berkley DB with new versions on GCC (v1.85/1.86 of Berkley DB specifically). They basically say "DON'T". I am in the process of re-writing the Berkley DB code with some modifications... right now for use in ypserv, but I think we really should consider replacing the libc stuff with them when complete. (They are much better in a number of ways... like in that it uses mmap(), thereby getting around almost all of the concurrent access problems (no more lseek()/read()/write() race conditions). It also pushes the buffer management problems back onto the VM layer w/o really consuming man memory for it (since the VM pages are backed by a real file). I also have implimented the R_CURSOR type for sequential access to HASH types, that alone helps a lot of things :) -- David Cross | email: crossd@cs.rpi.edu Lab Director | Rm: 308 Lally Hall Rensselaer Polytechnic Institute, | Ph: 518.276.2860 Department of Computer Science | Fax: 518.276.4033 I speak only for myself. | WinNT:Linux::Linux:FreeBSD To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message