See Jay Es Too Too Six

Winning benchmarks before it’s released: RHEL 5

nobody@vox.com(cjs226) — Sat, 17 Feb 2007 13:17:51 -0600

Keep in mind this is from Red Hat Magazine, but hey... ;-)

While many people are aware of SPEC benchmarks, they may be unfamiliar with the SPECompM and SPECompL series of benchmarks. These are used for characterizing the OpenMP shared memory performance in medium and large systems. OpenMP itself is a specification for compilers and libraries to make use of parallel directives. The types of problems this benchmark models are largely scientific in nature, covering everything from automobile crash simulation to ocean modeling to computational chemistry to genetics.

Many of these problems cannot be solved efficiently in a grid and still require shared memory multiprocessing resources. While Red Hat Enterprise Linux has been hugely successful in grid computing, its success in Symmetric Multiprocessing (SMP) has not been publicized as much.

SPEC OMP is of interest to High-performance computing (HPC) users, providing an objective and representative benchmark suite for measuring the performance of SMP systems. The focus is to deliver systems performance results appropriate for real scientific and engineering applications, so the benchmark places heavy demands on the processor, shared memory architecture, compiler and the OpenMP implementation. There are two OPM variants: OMPM2001 for 1-32 processor systems, and OMPL2001 for larger systems.

The Red Hat Enterprise Linux 5 SPECompM2001 result is the world record for a 16-way SMP configuration, and used an IBM POWER system that delivered a result of 45,895 - an incredible 78% faster than the previous Linux record holder. It is also a whole lot faster than the competing Unix results: 210% over Sun Solaris, 132% over HP/UX.

Many people are unaware of Red Hat Enterprise Linux’s ability to optimize large memory configurations, schedule across large numbers of CPU’s, and offer the compilers and libraries tuned to this problem space. So this benchmark is a terrific proof point for people who were waiting for Linux to mature in SMP space.

While commodity multiprocessors and server designs are cost optimized for the best price performance, large SMP systems are designed with performance as the prime goal. The ability to simulate an automobile crash in a computer, rather than building an actual model allows engineers to design a lightweight, yet strong car and iterate the design many times. This gives maximum safety and yet the light weight allows great fuel efficiency. Likewise the ability to model chemical properties in a computer allows tests for strength, toxicity, and cost of manufacture saves many times the costs of the computing systems.

This great result, beating all previous Linux results, shows the power and suitability of Red Hat Enterprise Linux in the scientific environment.

Read and post comments | Send to a friend

Solaris :: Create a Jumpstart server from ISO's

nobody@vox.com(cjs226) — Thu, 25 Jan 2007 17:11:19 -0600

Sun Solaris

Thanks to reborg from the unix.com forum for this!

The Solaris 9 and later ISO images are laid out differently than previous versions of ISO images for Solaris

To set up a JumpStart server using the Solaris 9 OS and later CD-ROMs, you must have access to both slice 0 and slice 1 on the CDs. If you try to use an ISO image that is loopback mounted using the "lofi" driver it cannot recognize that slice 1 exists and, cannot access it.

To fix the problem, split the contents of slice 1 into it's own image file and then mount this image separately using lofi.

The following procedure describes how to do this.
<ls -l sol-9*>
-rw-r--r-- 1 root root 576364544 Jan 1 11:16 sol-9-u1-sparc-v1.iso -rw-r--r-- 1 root root 291962880 Jan 1 21:42 sol-9-u1-sparc-v2.iso

This only applies to CD 1, all other isos can be mounted using lofiadm in the normal way.

- Get a copy of the VTOC from the ISO image:
- < dd if=sol-9-u1-sparc-v1.iso of=vtoc bs=512 count=1>

- Find out where Slice 1 starts in the image and how long it is. The starting cylinder for slice 1 is located at offset 452 (decimal) into the VTOC; the length in blocks is at offset 456, with both being 4 bytes long.
- <od -D -j 452 -N 8 < vtoc>
- Produces output similar to:
- 0000000 0000000888 0000546560 0000010

- Slice 1 starts on cylinder 888 and is 546,560 blocks long. CDs for the Solaris OS always have 640 blocks per cylinder, so you can find the starting block of slice 1 as follows:
- <echo 888*640 | bc>
    - Produces the following output:
      - 568320
        - So now you know s1 starts at block 568320 and is 546560 blocks long.

- Copy slice one into a separate file:
- <dd if=sol-9-u1-sparc-v1.iso of=sol-9-u1-sparc-v1-s1.iso bs=512 skip=568320 count=546560>

- Mount both slice 0 and slice 1 as follows:
- <mkdir –p /cd/s0>
- <mkdir /cd/s1>
- <lofiadm -a /path_to/sol-9-u1-sparc-v1.iso /dev/lofi/1>
- <lofiadm -a /path_to/sol-9-u1-sparc-v1-s1.iso /dev/lofi/2>

- When you mount slice 1, remember that it is a UFS partition, not HSFS as is usual on a CD-ROM:
- <mount -F hsfs -o ro /dev/lofi/1 /cd/s0>
- <mount -F ufs -o ro /dev/lofi/2 /cd/s1>

- Setup your jumpstart as normal:
- <cd /cd/s0/Solaris_9/Tools>
- <./setup_install_server /destination_dir>

Read and post comments | Send to a friend

Project 365 - Day 23 :: AIX...Who's your Daddy?!

nobody@vox.com(cjs226) — Tue, 23 Jan 2007 20:46:47 -0600

AIX...Who's your Daddy?!

After fighting AIX for days, <knocking on wood> I have conquered </knocking on wood> the OS that I have hated for years yet again.

Seriously, isn't it sad that the only positive I can think of for it is the running man that falls down when something fails in smit (GUI for sys admin work)? It's a tremendous PITA compared to Linux, Solaris and even (eek) Windows.

Read and post comments | Send to a friend

'New' CEOs and a new Sun-Intel partnership

nobody@vox.com(cjs226) — Tue, 23 Jan 2007 17:13:41 -0600

Sun

As with many couples in the early stages of courtship, some nice wine helped warm the mood between Sun Microsystems and Intel. Specifically, a bottle of Barolo at the swanky Delfina restaurant in San Francisco, where Sun Chief Executive Jonathan Schwartz met Intel CEO Paul Otellini.

"A really good bottle," Otellini recounted here Monday as the companies announced a new partnership to build Xeon-powered Sun servers and cooperate in software development. Sun previously had relied solely on Advanced Micro Devices' processors in its attempt to claim a place in the x86 server market it once shunned.

It was Schwartz who asked his opposite number on the first date. When he took over from Scott McNealy as Sun's CEO last April, the call to Otellini was the first Schwartz made, he said. "Surely there's more we can do together," he pitched the chipmaker chief, who had taken over the Intel reins from Craig Barrett less than a year earlier himself.

Read the rest at CNET.

Read and post comments | Send to a friend

Scripting past the NFS4 question after installing a Solaris zone

nobody@vox.com(cjs226) — Fri, 12 Jan 2007 12:43:26 -0600

Sun Solaris

Automating the NFS V4 configuration requires 2 steps. First, specify the value of NFSMAPID_DOMAIN in $zonepath/root/etc/default/nfs. Finally, you need to create a file called $zonepath/root/etc/.NFS4inst_state.domain to let sysidtool know that you have answered the question.

Read and post comments | Send to a friend

Solaris Zones :: Setting subnet mask in zones

nobody@vox.com(cjs226) — Thu, 11 Jan 2007 14:08:47 -0600

Sun Solaris

There isn't a netmask setting available when creating a zone and the /etc/netmasks file isn't consulted when plumbing up zone interfaces. This recipe describes how to force a netmask when creating a zone.

To configure an IP address of 10.1.1.12 with a class C subnet mask (255.255.255.0, a 24 bit mask) on physical interface bge0, use these commands when running zonecfg to create the zone (substitute your IP address, netmask length, and physical interface as appropriate):

zonecfg:testzone>add net
zonecfg:testzone:net>set physical=bge0
zonecfg:testzone:net>set address=10.1.1.12/24
zonecfg:testzone:net>end

This configuration is only necessary when using a non-standard netmask.

Read and post comments | Send to a friend

Using a different escape sequence to exit a Solaris zone

nobody@vox.com(cjs226) — Wed, 10 Jan 2007 20:43:49 -0600

Sun Solaris

To exit zlogin, use the escape sequence ~. on a new line (the tilde must follow a carriage return, not any other character). If this escape character is inappropriate (because you are connected using software like tip that also uses this escape sequence), you can specify a different escape character on the command line. For example, to use the @ symbol, use:

zlogin -C -e @ testzone

In this case, to exit the zlogin console session, type @. on a new line.

Read and post comments | Send to a friend

The future of the OS

nobody@vox.com(cjs226) — Sun, 17 Dec 2006 21:00:05 -0600

Virtualization

There's plenty of good reads on Scott Lowe's blog and I just finished one on his and other's thinking of the future of operating systems:

I do agree with these conclusions on at least one point: The general purpose operating system as we know it will cease to exist in the next 5 to 10 years, perhaps sooner. I do believe that the release of massive development projects such as Windows Vista won’t be the norm moving forward and that, in fact (as others have predicted as well), Windows Vista will be the last of its kind.

Notice I didn’t place Mac OS X in that list as well. Why? Because I think that Apple is capitalizing on an architecture and a convergence of technology that allows it to make Mac OS X into what Windows NT was supposed to be. (Go back and read the stories about the development of Windows NT and look at Dave Cutler’s vision for the operating system—an application environment-agnostic system in which OS/2, Windows, and POSIX-compliant applications could all run without modification.) Does that sound like anything else we have these days? With Mac OS X today, I can run native Macintosh applications, command-line UNIX applications (sometimes straight “out of the box”, sometimes needing a quick recompile), and X11-based applications. Add in something like WINE (the open source Win32 API implementation) and we gain the ability to run many (but not all) Windows applications. Add in a virtualization solution such as that created by VMware or Parallels and you gain the ability to run any Windows application.

Read the rest here.

Read and post comments | Send to a friend