From owner-freebsd-current Thu Jan 7 02:17:11 1999 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id CAA03324 for freebsd-current-outgoing; Thu, 7 Jan 1999 02:17:11 -0800 (PST) (envelope-from owner-freebsd-current@FreeBSD.ORG) Received: from caladan.tdx.co.uk (caladan.tdx.co.uk [195.188.177.4]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id CAA03319 for ; Thu, 7 Jan 1999 02:17:09 -0800 (PST) (envelope-from kpielorz@tdx.co.uk) Received: from tdx.co.uk (lorca-tx.tdx.co.uk [195.188.177.242]) by caladan.tdx.co.uk (8.9.2/8.9.2) with ESMTP id KAA81049 for ; Thu, 7 Jan 1999 10:16:39 GMT Message-ID: <3694895D.37DEEE92@tdx.co.uk> Date: Thu, 07 Jan 1999 10:15:57 +0000 From: Karl Pielorz Organization: TDX - The Digital eXchange X-Mailer: Mozilla 4.5 [en] (WinNT; I) X-Accept-Language: en MIME-Version: 1.0 To: current@FreeBSD.ORG Subject: Problem with -current, SMP and multiple 'nice' processes? Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Hi, I'm running -current as of a few days ago, and I've 'rediscovered' a long standing problem I've had... I run the RC5 client on my machine (OK, it's probably not that a 'technically correct' program). The machine is a dual PPro-200 (256Mb RAM), running an SMP a.out kernel... I've noticed this in the past, but thought it might have been 'fixed' or changed by now... If I run the client once (the client warns threads under FreeBSD aren't SMP aware etc. - so I tell it I have 1 CPU, and follow their other instructions {keep reading}).. With this 1 client running at a 'nice' priority level, the load average goes up to 1.0 - and everything is fine... If I run the client again {as per RC5's instructions for SMP FreeBSD machines} the load average goes up to 2.0 (as I'd guess you'd expect), and the machine performs like a dog... I have samba, sendmail + a few other bits & pieces running on the machine... Samba's performance drops from just under 5Mb's a second (via a 100Mb FDX link) to around 30-40k/sec). Everything else on the machine just 'judders' along very slowly... I've noticed this happening when I've been running other 'compute' intensive apps as well (I have a number of home-brew number crunching programs as well) - no amount of fiddling with 'nice' and priority levels seems to work... Is this just a 'symptom' of -current's present handling of SMP? - I know the machine will bottleneck around the kernel, but none of the programs I'm running are I/O or Kernel bound... If the concensus is that it's RC5's client that's 'bad' in someway, I might go on to write some more 'controlled' test programs etc. -Kp To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message