With the economy in crisis, IT departments worldwide are striving to reduce budgets and ‘green’ their IT architectures. With Red Hat Enterprise Linux 5, and the numerous “Green IT” features being added to Red Hat Enterprise Linux 6, Red Hat offers a highly power-efficient environment that aims to be the most ecologically friendly operating system platform on the market. Red Hat Enterprise Linux was named the “Greenest Operating System” in 2008 by Network World, beating out Microsoft Windows Server and SUSE Enterprise Linux, and was awarded the “Best Greener IT Product” at the 2009 SYS-CON Virtualization Conference & Expo Awards. Many of our customers, like Bank of New Zealand, have also made significant strides in greening their organization’s IT architectures with help from Red Hat Enterprise Linux.
To build on this leadership position, Red Hat has invested considerable resources during the development of Red Hat Enterprise Linux 6 to enhance and develop new power management technologies. Red Hat Enterprise Linux 6 is currently in Beta and is available for download. This blog provides a brief overview of the green improvements that are targeted for the upcoming release:
The new “tickless” kernel, in combination with numerous user-mode enhancements, previewed in recent Fedora releases, offers the possibility to reduce the number of wakeups per second for modern CPUs from 1024 (the ‘tick’ in Red Hat Enterprise Linux 5) to typically less than 30. Using well-documented tools, users may be able to lower this even more by manual tuning. This reduction makes it possible to quiesce the processor more effectively when the application load is low, thereby reducing power consumption.
Active State Power Management (ASPM)
ASPM is designed to reduce the power consumption on inactive PCI Express lanes. This feature is expected to be available for systems with relatively new hardware and firmware support. The implementation introduces a very small delay when an inactive PCI lane returns to an active state, but our internal testing has shown that the additional latency is negligible, so we believe that it is a worthwhile tradeoff.
Aggressive Link Power Management (ALPM)
The idea behind ALPM for SATA AHCI (Advanced Host Controller Interface) is to move a SATA (Serial ATA) link to a very low power mode when no I/O is pending. The controller is designed to automatically return the link back into an active power state as soon as an I/O is pending again. Savings of between 0.5 and 1.5 watts per SATA port can be expected. There are three different settings for ALPM. The default for non-hotplugable ports is the medium power setting, which seems to provide a good compromise between power saving and performance. More aggressive power saving can cause noticeable performance degradation, so while suitable for some custom environments, is not enabled by default.
Relatime drive access optimization
Relatime drive access optimization is a new method for avoiding metadata write operations on normal filesystem read accesses. In order to achieve this, the access time information is cached instead of being written immediately to a disk. The net result is that the speed of many operations, such as booting, is improved. The feature also results in increased link idle times and reduced disk spin time. Depending on the disks being used and the specific use case, this feature has the potential to save up to 2W per disk.
Enhanced graphics power management
Enhanced graphics power management is another area where Red Hat works continually to improve the abilities of the graphics drivers provided by different vendors. Depending on the hardware, our drivers offer the ability to do LVDS (low voltage differential signaling) reclocking, GPU reclocking and even complete GPU powerdowns.
Combined with relatime drive access optimization, the operating system has been carefully audited, using new SystemTap scripts, to identify applications that were doing unnecessary disk and network I/O. This iterative process is used to continuously optimize applications, reducing their I/O footprints and power demands.
Starting with Fedora 11, a new system service called Tuned was introduced. Briefly, this is a dynamic adaptive system daemon with a flexible plugin mechanism. As of today, monitoring and tuning plugins exist for ATA (Advanced Technology Attachment) hard-disk subsystems and Ethernet devices, and are available to measure CPU latency using PM-QOS (Power Management – Quality of Service). For Red Hat Enterprise Linux 6, this feature is expected to be combined with the ktune service, which provides an easier mechanism for static system tuning. We also anticipate that user definable and predefined profiles for various use cases, with an easy to use CLI, will be added.
Battery Life ToolKit (BLTK)
To analyze the impact of power management changes and enhancements, Red Hat engineers use a framework to generate reproducible workloads. The BLTK provides this, with several real life scenarios, ranging from a simple workload using a Firefox page reader to a more complex workload using OpenOffice writer and spreadsheet applications.
With all previous Red Hat Enterprise Linux releases, we have incorporated a number of partner-requested features to help the platform support the latest hardware capabilities. With Red Hat Enterprise Linux 6, we anticipate providing support for a number of partner-requested IT features designed to reduce energy use, such as HP Dynamic Power Capping and the Intel Node Manager. This type of feature support reflects the close working relationship that Red Hat seeks to maintain with its leading partners.
Virtualization and cloud computing
Red Hat Enterprise Linux continues to include improvements to its integrated virtualization technology. Virtualization provides obvious green IT benefits with the ability to reduce hardware sprawl and thereby reduce power and cooling costs. Typically, virtualization also provides the underlying foundation for cloud computing. With the ability to use only the software you consume, on-demand, cloud computing also offers enterprises the potential to make their IT infrastructure more energy efficient.
As you can see, Red Hat has put extensive resources into providing support for the latest power management capabilities, aiming to provide more energy efficient computing to our customers by both improving existing features and implementing new features. In many cases, these are being incorporated into higher level applications with the goal of making power management mechanisms more accessible and improve the user experience when using them.
In addition, guides, how-to’s and other documentation designed to make it easier to use the new power management features are available. And, of course, when the newest updates to Red Hat Enterprise Linux are released, we intend to provide additional details on power management features.
In summary, Red Hat is working toward the delivery of power management capabilities that are coherent, complete and appropriate for day-to-day use. Significant improvements have been made in the last few years, but there is always more to do. One of Red Hat’s goals is to ensure that customers can rely on Red Hat Enterprise Linux 6 to provide one of the world’s most energy-efficient computing platforms.