From owner-freebsd-questions Sat Mar 16 5:57:15 2002 Delivered-To: freebsd-questions@freebsd.org Received: from aasp.net (76-5.aasp.net [140.186.76.5]) by hub.freebsd.org (Postfix) with ESMTP id 3E81137B41A for ; Sat, 16 Mar 2002 05:57:06 -0800 (PST) Received: (apparently) from home98 ([141.154.9.189]) by aasp.net with Microsoft SMTPSVC(5.5.1877.197.19); Sat, 16 Mar 2002 09:02:45 -0500 Message-ID: <008101c1ccf2$dc569be0$1800a8c0@verizon.net> From: "Jules Gilbert" To: Subject: 3GB address space for user app's with FreeBSD 4.5 Date: Sat, 16 Mar 2002 08:59:55 -0500 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_007E_01C1CCC8.F16777E0" X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2600.0000 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000 Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG This is a multi-part message in MIME format. ------=_NextPart_000_007E_01C1CCC8.F16777E0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable I am running FreeBSD 4.5. Out of the box, I can malloc almost 512M. Nice for starter's, but not = nearly enough. To overcome this (in earlier editions of FreeBSD), I set = MAXDSIZ and DFLDSIZ to 1.5GB, giving me an effective 1GB of malloc'able = space. But I've made some changes to my application and I need to go higher, at = least to 2GB, and I would not be surprised in six months to be at 2.5GB. I have heard that 4.5 includes a feature to allow the user to increase = his available address space. I need to know more. I saw a reference to = an export'able variable. Is this right? What is the name of this = variable? Where is it documented? My application involves large-scale data mining, and I just upgraded to = 1.5GB per box. So, ideas anyone? Sincerely, Jules Gilbert PS1: When I set the MAXDSIZ and DFLDSIZ above, say 1700M, the machines = ocasionally issue double interrupts, which causes a third interrupt, = which in X86 is equivalent to a hard reset. Not very convienient! PS2: Years ago (until about '95) I ran Slackware Linux. Tried it again = for this application. It's changed, and today the Linux'es I tried were = bloated and performed very poorly compared to FreeBSD. Sure, not as = many applications run 'out of the box' as with Linux, but if you can use = a C compiler and want performance, the BSD's are the right choice -- = let's make the address space adaquate, too! PS3: What is really needed is a port of FreeBSD designed JUST FOR = cluster box'es. The machines I'm using are part of a small cluster = (each with 1.5GB and two disks). No terminals, no X sessions, no serial = com lines. Just an ethernet card and horsepower. Putting FreeBSD on = these boxes was incredibly difficult -- we wound up dup'ing disks. A = floppy based install via NFS would have been great. But my bigger point = is that no one is providing support for clustered boxes. ------=_NextPart_000_007E_01C1CCC8.F16777E0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
I am running FreeBSD 4.5.
 
Out of the box, I can malloc almost = 512M. =20 Nice for starter's, but not nearly enough.  To overcome this (in = earlier=20 editions of FreeBSD), I set MAXDSIZ and DFLDSIZ to 1.5GB, giving me an = effective=20 1GB of malloc'able space.
 
But I've made some changes to my = application and I=20 need to go higher, at least to 2GB, and I would not be surprised in six = months=20 to be at 2.5GB.
 
I have heard that 4.5 includes a = feature to allow=20 the user to increase his available address space.  I need to know=20 more.  I saw a reference to an export'able variable.  Is this=20 right?  What is the name of this variable?  Where is it=20 documented?
 
My application involves large-scale = data mining,=20 and I just upgraded to 1.5GB per box.
 
So, ideas anyone?
 
Sincerely,
Jules Gilbert
 
 
PS1:  When I set the MAXDSIZ and = DFLDSIZ=20 above, say 1700M, the machines ocasionally issue double interrupts, = which causes=20 a third interrupt, which in X86 is equivalent to a hard reset.  Not = very=20 convienient!
 
PS2:  Years ago (until about '95) = I ran=20 Slackware Linux.  Tried it again for this application.  It's = changed,=20 and today the Linux'es I tried were bloated and performed very poorly = compared=20 to FreeBSD.  Sure, not as many applications run 'out of the box' as = with=20 Linux, but if you can use a C compiler and want performance, the BSD's = are the=20 right choice -- let's make the address space adaquate, too!
 
PS3:  What is really needed is a = port of=20 FreeBSD designed JUST FOR cluster box'es.  The machines I'm using = are part=20 of a small cluster (each with 1.5GB and two disks).  No terminals, = no X=20 sessions, no serial com lines.  Just an ethernet card and = horsepower. =20 Putting FreeBSD on these boxes was incredibly difficult -- we wound up = dup'ing=20 disks.  A floppy based install via NFS would have been great.  = But my=20 bigger point is that no one is providing support for clustered=20 boxes.
 
------=_NextPart_000_007E_01C1CCC8.F16777E0-- To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message