Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 06 May 2003 16:06:24 +0200
From:      David Landgren <david@landgren.net>
To:        freebsd-current@freebsd.org
Subject:   Options available for using >4Gb RAM on x86
Message-ID:  <3EB7C160.207@landgren.net>

next in thread | raw e-mail | index | archive | help
List,

I saw Jake Burkholder's plea posted to /. a while back for people with 
large amounts of RAM to try out PAE support.

http://docs.freebsd.org/cgi/getmsg.cgi?fetch=670923+0+archive/2003/freebsd-current/20030413.freebsd-current

I'm looking at replacing my Squid proxy hardware in the next month or 
two. I'll be deploying a single processor (+HTT) HP DL-380 G3 with 6Gb 
RAM and two RAID 1 volumes.

This uses (at least in 4.x) the bge network driver and the ciss SCSI 
driver. Is this even reasonable to consider right now or are they 
known to be completely untested?

The current proxy supports over 500 users performing about 15 requests 
per second. This is hitting the limit of the IDE disk in the current 
machine. (systat -vmstat routinely shows the disk being 100% busy).

To me it seems like a good way of exercising both the network and disk 
subsystems. I realise that I may be gambling here, but I am prepared 
to fall back to the old machine the time it takes to revert to 4.8 if 
I had to, but then I'd be limited to 4Gb, which would be a bit sad.

Then again if 6Gb and/or the application isn't enough to contribute a 
useful datapoint then I'll probably specify the machine for 4Gb only 
and be done with it.

Thanks for any clues I can use (especially "I have the exact same 
setup and it works fine" :)

David Landgren



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3EB7C160.207>