OS Research News

Doctor Bletsch

Congratulations! Tyler Bletsch successfully defended his PhD dissertation. He is now entitled to all the rights, privileges, and honor thereunto appertaining.

and there was much rejoicing

JOP paper in ASIACCS 2011

Congratulations, Tyler.

"Jump-Oriented Programming: A New Class of Code-Reuse Attack" by Tyler Bletsch, Xuxian Jiang, Vince Freeh and Zhenkai Liang was accepted to ASIACCS. Acceptance rate is 16% (35/217).

Google Apps Engine outage

Read about google app engine failure here.

SNIA IOTTA Repository offers boatload of real IO traces

The SNIA IOTTA Repository offers a variety of real large-scale IO traces from a number of environments. These will prove invaluable to storage simulation or validation experiments.

Boot or install any Linux from the network with boot.kernel.org

My last post was on gPXE (quite a while ago), and I talked about all the amazing possibilities that come from a network boot-loader with that much intelligence.

It looks like these benefits are starting to be realized in a big way. Recently, kernel.org (which hosts the Linux kernel, as well as a ton of mirrors and other projects) released boot.kernel.org. This service has you download a tiny (~500kB) image for USB, CD, or floppy. Stick it on a device and boot from it and you'll have an easy-access menu of various Linux installers, Linux live environments, and test tools.

It's still rough around the edges, and not all the installers work quite right, but it's moving along, and shows a lot of ingenuity. The various environments leverage HTTP, HTTPFS (root filesystem over HTTP), and/or iSCSI.

As an aside, if you want to use their USB image, but still have your stick be able to hold data, it's easy. Just (1) copy existing stuff off, (2) dd the image to the stick as directed on the page, and (3) add a FAT partition back to the stick. Step 3 isn't hard:


$ wget http://boot.kernel.org/gpxe_images/gpxe.usb
...
$ sudo mount /dev/sdd1 /mnt
$ mkdir backup
$ cp -r /mnt/* backup
$ sudo umount /mnt
$ sudo dd if=gpxe.usb of=/dev/sdd
435+0 records in
435+0 records out
222720 bytes (223 kB) copied, 0.159969 s, 1.4 MB/s
$ sudo fdisk /dev/sdd

Command (m for help): p

Disk /dev/sdd: 2063 MB, 2063597056 bytes
64 heads, 32 sectors/track, 1967 cylinders
Units = cylinders of 2048 * 512 = 1048576 bytes
Disk identifier: 0x00000000

   Device Boot      Start         End      Blocks   Id  System
/dev/sdd4   *           1           2        2032   eb  BeOS fs

Command (m for help): n
Command action
   e   extended
   p   primary partition (1-4)
p
Partition number (1-4): 1
First cylinder (3-1967, default 3): (nothing)
Using default value 3
Last cylinder or +size or +sizeM or +sizeK (3-1967, default 1967): (nothing)
Using default value 1967

Command (m for help): t
Partition number (1-4): 1
Hex code (type L to list codes): b
Changed system type of partition 1 to b (W95 FAT32)

Command (m for help): p

Disk /dev/sdd: 2063 MB, 2063597056 bytes
64 heads, 32 sectors/track, 1967 cylinders
Units = cylinders of 2048 * 512 = 1048576 bytes
Disk identifier: 0x00000000

   Device Boot      Start         End      Blocks   Id  System
/dev/sdd1               3        1967     2012160    b  W95 FAT32
/dev/sdd4   *           1           2        2032   eb  BeOS fs

Partition table entries are not in disk order

Command (m for help): w
The partition table has been altered!
...
Syncing disks.

$ sudo mount /dev/sdd1 /mnt
$ cp -r backup/* /mnt
$ sudo umount /mnt

gPXE: Modernized Ethernet boot

I just read up on gPXE, which is an open source alternative to proprietary PXE bootloaders. It supports booting a variety of environments and can be flashed onto NICs or chainloaded from regular PXE.

What struck me was this: it supports HTTP download of both gPXE scripts and binary images. This means that you could set up a web server that gives out gPXE scripts based on a PHP script or other server-side code. The end result is the ability to build a cluster whose nodes boot based on the control of a simple web server. If applied to something like the NCSU VCL, it eliminates the need for disk image transfers: nodes could be configured to run a certain environment by flipping one bit on a web server instead of pushing gigabytes of disk image data.

Furthermore, this isn't just restricted to sickly old NFS boot! gPXE supports software iSCSI boot, so all modern Linux and Windows OSes can get in on the act with just a small amount of configuration.

If I had more time, I'd love to apply this kind of deployment on our research clusters. Just putting the selection of Linux kernels in a centralized place would be a big help, even if we still mount the local disk as root!

Into the cloud: a conversation with Russ Daniels

Ars Technica is carrying an interview with Russ Daniels, HP's CTO and VP of Cloud Services Strategy. He discusses a practical definition for cloud computing and distinguishes it from utility computing.

Part 1
Part 2

Arguments against auto-scaling compute capacity in cloud computing

We've been interested in applying power-aware techniques to virtualized compute grids, including so-called "cloud computing" grids. A key part of this work is dynamically scaling the amount of compute capacity based on load, which in some cases can be done automatically.

This article presents an opposing viewpoint and points out some ways in which automatically-driven dynamic resizing of compute resources might be a bad thing. I think the author goes a bit to far in throwing out automatic scaling entirely, but he does raise several valid points, especially with respect to security and the possibility of DDOS attacks.

Overall benefits of CPU power savings in a data center

I found an article that shows interesting analysis of power savings in a data center. In a 5,000-square-foot data center, a 1-watt reduction at the server-component level (processor, memory, hard disk, etc.) results in an additional 1.84-watt savings in the power supply, power distribution system, UPS system, cooling system, etc. Consequently, every watt of savings that can be achieved on the processor level creates approximately 2.84 watts of savings for the overall facility.

Flash based storage not so energy efficient?

Tom's Hardware tested the power and performance of current Solid State Disks (SSDs) both in terms of performance and power. While the performance numbers were as expected, it turns out that the flash based drives' energy utilization is no better than a traditional 7200 RPM hard disk for a practical workload based on the MobileMark benchmark. The authors contend that this is because hard disks reach their maximum power draw only when seeking, whereas flash storage uses full power during any IO activity.

Ubuntu 8.04 released

Ubuntu 8.04 was released earlier this week. You can get it from one of the official mirrors or from our unofficial mirror. This is a "Long Term Support" (LTS) release, meaning that updates will be available for 3 years on the desktop and 5 years for servers.

AMD Barcelona architecture rundown

The Barcelona (aka "K10") microarchitecture is the latest design from AMD for both the server and desktop markets. The Phenom is the quad-core desktop variant, the Athlon X2 series includes the dual-core variant, and the 23xx and 83xx Opterons are the quad-core server varient.

The key changes over the previous line are covered in brief here and in greater detail here. Most of the interesting features require the use of an upgraded CPU socket denoted by a "+" (e.g. Socket AM2+ or Socket F+), though the CPU will work in non-plus sockets on current motherboards. Some of the "plus socket" features are:

• Separated voltage planes allow the CPU to have a different voltage/frequency for each core and the northbridge.

• HyperTransport 3.0, allowing greater bus bandwidth, including support for DDR2-1066.

In addition, the Barcelona introduces a shared L3 cache, which should have a major impact on HPC applications.

One major issue, however, is an L3 TLB bug present in the first generation of this architecture. This problem can be solved by disabling part of the L3 TLB system in the BIOS or via software (with a 10% performance penalty), or using a unique Linux patch to route around the problem with limited slowdown (but the patch is not intended for production use). See the Phenom wikipedia article for details.

In short, while Intel retains the upper hand in horsepower now, the AMD Barcelona design seems to sport many of the features predicted for future system design.

More information:

Wikipedia's Barcelona article covers the architecture in depth.

Anandtech benchmarking puts the chip through its paces.

To find a Barcelona-based chip, see Wikipedia:

• Phenom quad-cores
• Barcelona-based dual-core Athlons (scroll to "Phenom based")
• Barcelona-based quad-core Opterons (23xx and 83xx)

"The Coming Utility Computing Revolution"

This recent "Innovations" article highlights "utility computing", the idea that virtualization, shared storage, and other technologies will come together to commoditize business computing.

While I agree with the general idea, the author predicts that this will marginalize IT as a field, which seems counter-intuitive. While this kind of computing does allow fewer people to manage more systems, it does make that management that much more complicated. Further, IT has always been about helping users as much as maintaining infrastructure. So I don't see the general IT realm getting eaten by other fields, but rather splintering into specialists in networking, storage, (virtual) system administration, support, etc.

Finally, I found this quote pretty funny:

Teenagers entering higher education today are already skilled at building personal application spaces on Facebook using software modules. It’s a small step to apply those principles to business applications.

"A small step" to go from facebook to a crucial business application? Seems unlikely.

Comparison of memory model between Xen and OpenVZ

The article explains the difference in detail. Now, I can understand why it's too difficult to run Java Web server on OpenVZ for evaluating TPCW benchmarks.

Power management research at Berkeley

There is some good information about adaptive power management at Berkeley. It appears to be a position statement/paper in construction arguing for power management in data centers.

Ubuntu 7.10 released

Ubuntu 7.10 was released today. You can get it from one of the official mirrors or from our unofficial mirror. Notable new features include:

- Improved support for multiple monitors, rotatable monitors, etc.
- Plug and play printer detection
- NTFS writing support out-of-box
- Kernel "dynamic ticks", meaning that the kernel doesn't have to wake up 1000 times per second any more, thus saving power.
- Full or partial hard disk encryption selectable at install time

Okay, this still isn't "research" news, but I like it.

EPA Data Center Report

The EPA has issued its report on data center energy consumption.

Citrix acquires XenSource, VMware goes public

By way of Ars Technica, we find that Citrix (whose software provides virtual Windows desktop environments accessible anywhere, among other things) has purchased XenSource, the commercial arm of the Xen hypervisor virtualization project. Hopefully they'll make it less of a pain to use and deploy Xen.

Also, VMware went public last Tuesday with a successful IPO. Hopefully they'll keep putting out non-commercial freeware like VMware server and player. Maybe we'll some day seen a free (but likely crippled) version of ESX?

Linux Gains Two New Virtualization Solutions

By way of slashdot, we find that the Linux kernel (2.6.23 and up) now sports three virtualization techniques out-of-box: KVM, Xen (just merged), and Lguest (also recently merged).

Lguest in particular looks interesting, as it doesn't require virtualization hardware support (like KVM), but is as simple as a single modprobe (as opposed to the Xen behemoth). Performance isn't too great right now, though (-30%).

Power Plays: How power consumption will shape the future of computing

This Ars Technica article spotlights the development of power-aware technologies at the chip, system, network, and data center levels. It analyzes recent developments in terms of granularity, i.e. the frequency of reaction. Overall, a well written article.