Re: Best 10 Gig PCIe for R900?

2010-10-08 Thread Peter Kjellstrom
On Friday 08 October 2010, David Hubbard wrote:
 Hi all, I've got an R900 with four six-core cpu's that
 will be doing Symantec NetBackup backing up and
 de-duplication duties over ten gig via a Cisco 4900M
 so it's going to need to be a fiber card, short range.

 Any recommendations on best NIC for the job, i.e.
 best throughput and lowest resource utilization?  OS
 is RHEL 5 and ideally I'd like to stick with the
 built in drivers but if necessary I can replace
 them, I just hate dealing with 3rd-party drivers and
 kernel patches and rebooting with no networking, etc.

I've recently tested three different 10G eth NICs on a similar platform using 
CentOS-5.5(x86_64). Here are the NICs with driver comments:

 Intel X520-da2: driver: ixgbe
 Built in driver good, no upgrade req.

 Emulex OneConnect: driver: be2net
 Built in driver ok, upgraded driver better (performance)

 NetXen NX3031: driver: nx_nic
 No built in driver

All of these NICs performed well and =2 tcp streams reached wire speed 
without any tcp-tuning. I've not done extensive testing on CPU-consumption 
but most servers today have cores to spare, no?

A 4th NIC worth considering IMO is the ConnectX from Mellanox.

/Peter


signature.asc
Description: This is a digitally signed message part.
___
Linux-PowerEdge mailing list
Linux-PowerEdge@dell.com
https://lists.us.dell.com/mailman/listinfo/linux-poweredge
Please read the FAQ at http://lists.us.dell.com/faq

Re: Disk performance drop in new RHEL 5 kernel?

2010-08-27 Thread Peter Kjellstrom
On Thursday 26 August 2010, Drew Weaver wrote:
 Hi there, I am testing some R510s /w PERC H700 that I am going to be using
 for storage systems.

 I am using a very basic benchmark (hdparm -tT /dev/sdb) to get a relative
 performance number between different disks (it doesn't have to be real
 specific, I just need to see percentages based on different physical
 media).

 Anyway, I noticed that if I run hdparm -tT /dev/sdb on RHEL 5.4 (2.6.18-164
 (megasas 04-08-RH2)) in an average of 10 runs I get 825MB/sec, however if I
 update the OS to 5.5 2.6.18-194.11.1 (megasas 04.17-RH1) in an average of
 10 runs I get 561MB/sec. I am trying to figure out if this is a real
 performance decrease or if there were changes made in the new kernel or
 driver that would cause hdparm to have an issue getting the disks to work
 hard.

 If I reboot with the older kernel it is always around 825MB, I can
 reproduce this over and over.

 I will also note that in Fedora 13 it is also around 825MB/sec.

 Has anyone seen this, any thoughts?

Yes, read and write some real files to see if this is an actual drop in 
performance. hdparm is not a good benchmark.

/Peter


signature.asc
Description: This is a digitally signed message part.
___
Linux-PowerEdge mailing list
Linux-PowerEdge@dell.com
https://lists.us.dell.com/mailman/listinfo/linux-poweredge
Please read the FAQ at http://lists.us.dell.com/faq

Re: H200

2010-08-10 Thread Peter Kjellstrom
On Thursday 29 July 2010, Drew Weaver wrote:
 Has anyone done any testing for the H200 for various Linux distributions
 yet?

We've done some testing on 64-bit CentOS-5.4 and 5.5. It just works (as a 
sas-controller, raid is only 1/0) and we got quite good performance (aggreate 
software raid0 12 drives r/w: 1000 MB/s / 750 MB/s).

/Peter

 Also does anyone have any performance numbers between the H200 and the SAS
 6iR?

 The SAS6i is well supported by pretty much every Linux distro (even old
 versions) but if the performance of the h200 is better it may be worth it
 to switch.


signature.asc
Description: This is a digitally signed message part.
___
Linux-PowerEdge mailing list
Linux-PowerEdge@dell.com
https://lists.us.dell.com/mailman/listinfo/linux-poweredge
Please read the FAQ at http://lists.us.dell.com/faq

Re: Hot disk change.

2010-04-19 Thread Peter Kjellstrom
On Monday 19 April 2010, Fabio Catunda wrote:
 Hi,

You're missing two steps below.

 I have a PowerEdge 1950 with a PERC 5/i Integrated controller on
 RAID-1 with two 232.25GB disks.

 I need some more disk space, so I bought two 2TB disks.

 What I would like to know is if it's possible to replace one HD at a
 time, trusting the controller to sync all data with no loss!?

Assuming the rebuilds go well and you now have your raid1 on two 2T drives 
you'll need to expand the raid1 (if possible).

 The OS is a Debian Linux with ext3 fs, so after changing both disks I
 could use resize2fs to use the extra capacity.

Before resizing the filesystem you'll have to take care of the layer beneath 
(partition table, lvm or similar).

 Anyone already tried something like that?

 Any tip is appreciated.

 Thanks,

 Fábio Catunda.


signature.asc
Description: This is a digitally signed message part.
___
Linux-PowerEdge mailing list
Linux-PowerEdge@dell.com
https://lists.us.dell.com/mailman/listinfo/linux-poweredge
Please read the FAQ at http://lists.us.dell.com/faq

Re: RHEL ES 3 Kernel Upgrade Issue

2010-04-14 Thread Peter Kjellstrom
On Wednesday 14 April 2010, Murphy, Brian S CTR USAF AFSPC 83 NOS/Det 4 wrote:
 Ah - http://mirror.centos.org/centos/3/updates/i386/RPMS/ has

 kernel-2.4.21-63.EL.i686.rpm
 and
 kernel-smp-2.4.21-63.EL.i686.rpm

 both dated 4 Nov 2009 - looks good.

 Is there some way to know how that translates to kernel.org versions (or
 are they patches to 2.4.21 put out to fix CVE issues)?

Vendor versions like RHEL/CentOS/SLES/whatever can't really be translated to 
kernel.org versions. From a base-line perspective 2.4.21-63.EL may be near 
2.4.21.x but some parts are backported from much higher versions. From a 
security perspective you either read the RHSA (RedHat Security Announcements) 
and/or the rpm changelog (rpm -qp --changelog kernel...rpm).

/Peter

 Or how other CentOS RPMs there translate to vendor versions (BIND shows
 9.2.4-25, but ISC's named versions are at 9.4.x and 9.6.x, etc.)?

 Thanks!

 Brian


signature.asc
Description: This is a digitally signed message part.
___
Linux-PowerEdge mailing list
Linux-PowerEdge@dell.com
https://lists.us.dell.com/mailman/listinfo/linux-poweredge
Please read the FAQ at http://lists.us.dell.com/faq

Re: Third-party drives not permitted on Gen 11 servers

2010-02-10 Thread Peter Kjellstrom
On Tuesday 09 February 2010, William Warren wrote:
 On 2/9/2010 5:17 PM, howard_sho...@dell.com wrote:
  Thank you very much for your comments and feedback regarding exclusive
  use of Dell drives. It is common practice in enterprise storage solutions
  to limit drive support to only those drives which have been qualified by
  the vendor.  In the case of Dell's PERC RAID controllers, we began
  informing  customers when a non-Dell drive was detected with the
  introduction of PERC5 RAID controllers in early 2006. With the
  introduction of the PERC H700/H800 controllers, we began enabling only
  the use of Dell qualified drives.
 
  There are a number of benefits for using Dell qualified drives in
  particular ensuring a positive experience and protecting our data.
...
 This is common reasoning given for any vendor that starts practicing
 lock-in.  Dell has just gone down that road.  I'll either not buy Dell
 servers OR order them without your controllers and use some of my own.

If they'll allow you to use non-Dell controllers...

/Peter

 Over the years proprietary solutions are only cash cows and rarely if
 ever really live up to the claims put forward by the vendor.


signature.asc
Description: This is a digitally signed message part.
___
Linux-PowerEdge mailing list
Linux-PowerEdge@dell.com
https://lists.us.dell.com/mailman/listinfo/linux-poweredge
Please read the FAQ at http://lists.us.dell.com/faq

Re: Third-party drives not permitted on Gen 11 servers

2010-02-10 Thread Peter Kjellstrom
On Wednesday 10 February 2010, John Oliver wrote:
 On Wed, Feb 10, 2010 at 11:36:54AM -0500, J. Epperson wrote:
  And UPSs!  We must ensure that we have appropriately proprietarily
  conditioned power for our proprietary servers.  And no third party
  replacement batteries either.  Lord only knows what sort of corruption
  that could lead to.

 --
 And, for our own good, we must be restricted to using Dell-branded
 enclosures... we can't take the risk that the holes in a lesser product
 might be out-of-kilter, slightly tweaking the chassis, causing memory
 and cards to edge out of their slots.

 Better get Dell-branded network cables, too.  And Dell KVMs, Dell
 keyboards, Dell mice... Dell mousepads, too, just to absolutely ensure a
 positive computing experience...

Be careful what you wish for. The network cable-one is already here. 10G for 
shorter distances now is using so called twinax cables for SFP+. One feature 
of SFP+ (like other transiever slots) is that it identifies what you connect. 
Atleast one major network equipment vendor (and I'm guessing most other too) 
only allows its own cables.

/Peter


signature.asc
Description: This is a digitally signed message part.
___
Linux-PowerEdge mailing list
Linux-PowerEdge@dell.com
https://lists.us.dell.com/mailman/listinfo/linux-poweredge
Please read the FAQ at http://lists.us.dell.com/faq

Re: Lifecycle Controller not-so slick IMO.

2010-01-28 Thread Peter Kjellstrom
On Wednesday 27 January 2010, patrick_b...@dell.com wrote:
 By Console Redirection we mean serial port console redirection. The
 interface works fine over the iDRAC virtual KVM. This was decided since the
 Lifecycle Controller uses a graphical interface.

At first your description seemed great but as more information trickles in it 
seems less and less useful.

The idea to do firmware upgrades before OS is great, but feeding and 
triggering this has to be done in a scalable way.

Why not simply upload firmware blobs (or URLs to such) to the MC/DRAC which is 
then applied at next reboot (result logged to SEL).

Just my $0.02,
 Peter


signature.asc
Description: This is a digitally signed message part.
___
Linux-PowerEdge mailing list
Linux-PowerEdge@dell.com
https://lists.us.dell.com/mailman/listinfo/linux-poweredge
Please read the FAQ at http://lists.us.dell.com/faq

Re: issues with bnx2 - NETDEV WATCHDOG: eth0: transmit timed out

2009-12-15 Thread Peter Kjellstrom
  On our new R410, randomly when eth0 is pushing more than 30Mbps of
  traffic
  the interface sometimes times out and shuts off:
...
  This morning I found this post suggesting adding options bnx2
  disable_msi=1
  in /etc/modprobe.conf

Do report back and tell us if this fixed your specific situation.

/Peter

 Thanks,

 --
 pjf


signature.asc
Description: This is a digitally signed message part.
___
Linux-PowerEdge mailing list
Linux-PowerEdge@dell.com
https://lists.us.dell.com/mailman/listinfo/linux-poweredge
Please read the FAQ at http://lists.us.dell.com/faq