From owner-freebsd-hackers Wed Mar 6 15:45:07 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id PAA10606 for hackers-outgoing; Wed, 6 Mar 1996 15:45:07 -0800 (PST) Received: from tombstone.sunrem.com (tombstone.sunrem.com [199.104.90.54]) by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id PAA10598 for ; Wed, 6 Mar 1996 15:45:04 -0800 (PST) Received: (from brandon@localhost) by tombstone.sunrem.com (8.6.12/8.6.12) id QAA05073; Wed, 6 Mar 1996 16:44:44 -0700 Date: Wed, 6 Mar 1996 16:44:44 -0700 (MST) From: Brandon Gillespie To: hackers@freebsd.org Subject: Re: Linux vs FreeBSD comparison - it's time, I think! In-Reply-To: <199603062206.RAA01997@etinc.com> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-hackers@freebsd.org Precedence: bulk > then things that you selected aren't the issue. But if you do network > testing with a card that has a particularly bad LINUX driver and a > very good FreeBSD driver, then the test is only valid for that one > particular card. The Linux people could do the same test with a > very good Linux driver and a buggy FreeBSD driver and get > opposite results. So what have you shown? > > For example BSDI with an Adaptec SCSI and an NE2000 card is not > a very good system, but with a buslogic VLB or EISA card (they've done > a lot of work with these cards) and an SMC ethernet they're very good. Which is why it is important to list all hardware specifications, and to get a wide variety of benchmark results. Yes, GenericOS-1 may work well with cardx, where GenericOS-2 does not, but if GenericOS-2 works well with 10 different other cards which are more common, then I'd say genericOS-2 has the advantage. The only way to find this is by getting many results. Why don't we come up with a 'form' which is filled out by a person performing a benchmark, which includes everything from I/O cards to seek times on drives (as well as the specific kernel version/config for each OS you benchmark on that hardware). From there, take the information from various benchmarks (such as lmbench) and collect them into a database. We should also get somebody to grab the linux FAQ for easy install instructions.. I can come up with a web page to handle submitting benchmarks if somebody can do something with the information (and if we can come up with a list of generic items to consider in a hardware config (i.e. i/o cards, drive specs, chipsets, etc). -Brandon Gillespie- BTW, brainstorming on specing hardware, have brand, model and other specs for all primary hardware pieces. Those would be ... processor, hard drive (spec the controller with each drive), i/o (other?), network, video, ...?