PowerEdge MegaRaid Linux large files

Steve_Boley@Dell.com Steve_Boley at Dell.com
Thu Aug 30 16:52:00 CDT 2001

You also have to remember that these 2 controllers (the perc2/sc and percII)
are both hampered in several ways apart from the later models of the ami
controllers.  The sc has no battery backup cache, 16mb edo simm, 33mhz
processor, and a plus is lvd speed 80mbs: percII is 33mhz, 16 and 32mb simm
cache, has battery cache, but is limited to uw scsi 40mbs.  There is
defenitely a point where these specs are a bottleneck to large file
transfers in system.  If transfer is involved with a database then is a
possibility that the cache on the card is flooded.  If you are running only
one logical drive then the following fix will kill your os system
performance, but if you have your os on a log drive and your data on another
then you can change the cache configuration on the logical drive to write
through instead of write back.

-----Original Message-----
From: Michael E Brown [mailto:michael_e_brown at dell.com]
Sent: Thursday, August 30, 2001 4:17 PM
To: linux-poweredge at exchange.dell.com
Cc: linux-poweredge at exchange.dell.com
Subject: Re: PowerEdge MegaRaid Linux large files

What you may be experiencing is a generic Linux kernel tuning issue.
Writes favor reads in the kernel I/O subsystem and can sometimes starve
other parts of the system from reads. Some tuning that will be present in
Red Hat Linux 7.2 when it comes out may fix it.


In particular, note the comment from Arjan:

------- Additional comments from arjanv at redhat.com 2001-08-30
Note, you can tune the maximum time reads wait for writes with the
/sbin/elvtune tool. Making reads have a lower latency will improve
interactive performance at the cost of raw throughput (and thus benchmark

Also, if this is a Perc or similar device; such a device can easily have
up to 100 Megabyte of IO "in flight" in the controller and then it takes a
while to complete new read requests.


On Tue, 28 Aug 2001, Ron Stanonik wrote:

> Hi,
> We bought a PowerEdge 6300 (year or two ago) with 1GB RAM,
> six 18GB disks, PERC2/SC (megaraid), and RedHat (6.X at the
> time, 7.1 now).
> The six disks were configured as one raid 5.
> If anyone writes a large file, any other writes hang until the
> large file finishes.  Since almost any activity creates files
> (email, /tmp, utmp, etc), all activity hangs until the large
> file finishes.  Of course the large file must be really large
> (>100GB) for the hang to be noticeable, but for us this happens
> often enough to be a problem.
> We've flashed the latest megaraid BIOS (3.13) and are running
> the (almost) latest megaraid driver (kernel 2.4.8, driver 1.17).
> If we configure the six disks as one raid 0, then the problem doesn't
> occur.  Bummer, we'd really like the protection of raid 5.
> We keep hoping that the next BIOS update or kernel version or driver
> version will fix it, but not yet.
> Wondering if you've encountered this, have any hints, etc.
> Thanks,
> Ron Stanonik
> stanonik at cogsci.ucsd.edu
> (858)822-2356
> _______________________________________________
> Linux-PowerEdge mailing list
> Linux-PowerEdge at dell.com
> http://lists.us.dell.com/mailman/listinfo/linux-poweredge

Michael Brown
Linux OS Development
Dell Computer Corp

  If each of us have one object, and we exchange them,
     then each of us still has one object.
  If each of us have one idea,   and we exchange them,
     then each of us now has two ideas.

Linux-PowerEdge mailing list
Linux-PowerEdge at dell.com

More information about the Linux-PowerEdge mailing list