Green computing on AIX and POWER
Definition and evolution: IBM green computing
You hear a lot about green computing, but what exactly is it, why is it important, and what is IBM’s involvement in green computing? In essence, green computing defines the environmentally responsible way of deploying computer-based resources and technology. But why should you care? The short answer is that as a responsible person or company whose clients are increasingly becoming more eco-friendly and responsible themselves, it is imperative that you start your efforts towards green computing for both indirect and direct reasons. The direct reasons involve the cost of doing business, and the indirect reasons relate to environmental, eco-friendly concerns.
The biggest challenges for implementing green computing revolve around the server, the rack, and the datacenter. IBM has made significant progress in all three areas, much of it around the POWER6. Through various studies, IDC has estimated that for every dollar spent on hardware, another 50 cents is actually spent on energy and that number is expected to increase to 71 cents by 2011. In a typical datacenter, 55% of energy goes to power and cooling, with the remainder going towards server management. In a society with dramatically rising energy costs, where levies are being imposed on carbon production in many countries, it has become a fiscal, legal, and moral imperative to look towards responsible methods of implementing computing technology. Make no mistake, financial penalties already exist in many countries and more are coming. The bottom line is, do you want to be (or work with) a proactive organization that thinks seriously about environmental concerns, or do you want to be reactive and take measures only when threatened with a loss of revenue?
IBM has long been on the forefront of responsible computing. It has had the size, wherewithal, and experience to help themselves and customers implement green computing. For example, between 1990 and 2005, long before green became sexy, IBM’s Global conservation efforts actually reduced carbon consumption by 40% of its 1990 emissions. Chip manufacturers have always been under a tremendous amount of pressure to developer faster, smaller transistors, which usually run hotter and consume more electricity. The rising temperatures add to the consumption of power, which in turn requires additional cooling. Rather than this decreasing the amount of iron needed in the datacenter, the amount of systems in the datacenter has actually increased through the years. This is because users have been taking advantage of these gains to drive applications and databases that require yet more power. Furthermore, some of these small chips allow manufacturers to produce smaller rackable systems that allow even greater computing power to be brought back to the datacenter, again, with even more demands on energy, power, and cooling. While virtualization has certainly helped cut down on the number of physical boxes necessary to support instances, it also has created new demands on IT to become that much more efficient in its ability to deploy new instances. The net is that the better the technology, the more in demand it is, not less. Studies have shown that even with virtualization, the number of servers in datacenters has increased 18%.
Project Big Green to the rescue
In May of 2007, IBM announced a one-billion dollar investment to increase energy efficiency within their own datacenters and the clients that use them. This defined “Project Big Green,” which targeted datacenters and included a team of engineers of over 800 energy engineers worldwide to work on the initiative. The plan ultimately was to reduce datacenter costs by an average of 42%. The five areas that would be covered included: diagnose (assessment of exiting faculties), build (plan build out of datacenters), virtualization (virtualize infrastructure), manage (using software to manage the process), and cooling (exploitation of liquid cooling resources). Active energy management, which allows for the capability to manage facilities, cooling, and electrical systems within the datacenter, is a big part of this initiative.
One of the primary drivers for Project Big Green was purely operational, meaning that customers had a hard time deploying new servers because of datacenter energy requirements and heating issues. Cost also became a huge factor, as increased reliance on the datacenter and the increase in data capacity also required substantial increase in cooling costs. The social and environmental factors also started coming into play. Without question, customers wanted to put in place more energy-efficient systems. This was in response to pressure from customers, shareholders, and even the government. With a new environmentally conscious administration, this pressure will only increase. IBM has even started a program with its business partners to help develop programs that will provide for more efficiency. As companies increasingly ask to become more eco-responsible, hardware vendors needs to react to that. IBM has been very proactive in this respect.
IBM just started a Big-Green service, where they will work with customers on the five datacenter areas the article already discussed: diagnosis, building, virtualization, managing, and cooling. An important part of this program is IBM’s mobile measurement technology, which can actually measure temperature in datacenters, looking for air leakage and other problems. Along with this, IBM will also provide a thermal analysis for high-density computing that provides additional options for savings.
POWER6
The POWER6 was built on IBM’s 65-nanometer process technology. In essence, it reduced the size of silicon, which allowed engineers to reduce overall processor consumption. At the same time, IBM also was able to increase its performance to 4.7 GHz while not increasing the energy costs at all. IBM did this by keeping the number of pipeline stages static. This is the number of operations that must be completed in a single cycle of clock time. Doubling the performance while using the same thermal envelope is perhaps the most impressive innovation of the POWER6. The design of the chip also itself supports the separation of circuits that have problems supporting low voltages. This further allows the actual power to be reduced on other circuits within the chip.
The clock itself can also be dynamically turned off when no work is being done and then turned back on again when needed. This is due to the PowerSave mode of the POWER6, which provides the means to drop voltage and frequency of the system, enabling power usage savings of approximately 30%. The POWER6 processor actually drops to 35 percent of power consumption when it is idle. Furthermore, memory can also be shut on and off as required to help balance server workloads, which have the effect of further decreasing energy consumption.
Other POWER6 innovations include:
* IBM’s POWER6 JS22 blade (The first POWER6, which was introduced in 2007). IBM’s BladeCenter® was equipped with a shared power infrastructure, which provided more efficient power supplies that could reach peak efficiency even with a minimal load, saving approximately 28% over other power supplies. IBM’s calibrated vector cooling technology further allowed the system to save money by providing for dual paths of air for each component, adding to the longevity of systems and reducing wasteful air movement.
* IBM’s Power 575. To increase the rack density, IBM created a new modular water cooling and distribution system. Part of this system includes a rear door heat exchanger that can extract 60% of the heat generated by the servers themselves. Next generation systems may even be able to extract 100% of server heat generation. Imagine an environment that does not need cooling at all!
* PowerVM™ virtualization front. Live partition mobility (a POWER6 virtualization innovation) provides for the ability of customers to save energy by relocating and consolidation active partitions, which can further free up resources. You can also power down entire servers that are not needed.
EnergyScale™
EnergyScale is another important innovation on POWER6, which provides for the ability to control power and cooling, allowing for better facilities planning, energy control, and systems availability. Among other features, it has a power-capping mode, allowing users to actually configure maximum levels of power access across systems. This feature provides datacenter operators with the ability to put caps on the amount of power that certain servers are using, ensuring that the system can never exceed a certain amount of power. By doing so, the operator can free up power across the datacenter for servers that really need the power. This feature is available on all POWER6 systems.
Some other important features of EnergyScale include:
* Power trending. This provides for the collection of power consumption data in real time. The usage itself is displayed by Active Energy Manager, now a component of IBM Systems Director. One can monitor consumption daily, weekly, or even monthly. It’s important to note that for the Power 560 and 570 models, an Intelligent Power Distribution Unit is required, which provides the power to the computer. The other systems have built-in power meters and do not require add-ons.
* Thermal reporting. This innovation allows for the lowering of processor frequency and voltage by a fixed amount. This has the effect of reducing power consumption on the system while still delivering peak performance. Active energy manager actually is the mechanism that enables and disables the powersaver mode. An example of where this can be used is on weekends where workload levels can shift down dramatically.
* Energy-optimized fan control. This works with the firmware on POWER6 systems and allows for an adjustment in fan speed based upon power consumption and temperature, among other variables.
* NAP core. The processor has a low power mode that stops processor execution when there is no work going on. This mode is called nap mode. The way it works is that the OS detects when a thread is idle and then gives control of the hardware back to the hypervisor, which puts the thread into nap mode.
* EnergyScale for I/O. Power Systems can now power-off pluggable, PCI adapter slots that are not being utilized, saving approximately 14 watts per slot. The slot is powered off as soon as it is removed from the partion. All POWER6 processor systems support this mode. This is available for hot-pluggable PCI slots only.
* Processor folding. This is a technique that dynamically adjusts the amount of processors dispatched, which try to match the workload. In essence, it increases the energy savings of the system during low to moderate workloads. The tougher the workload, the greater the amount of processors become available. In a way, this is similar to using logical partitioning changes but with much greater efficiency, because it does not impact the configuration of the partition nor processor utilization of the partition itself.
Active Energy Manager
Active Energy Manager (AEM) is the next generation product to PowerExecutive, which was previously only available on x86 systems. Driven by IBM System Director, it allows you to monitor and manage server requirements. When utilized correctly, AEM will allow you to lower power usage on servers, use less power and cooling on the systems, determine power usage for rack components, and perform power/thermal trending and analysis of energy consumption to help with future capacity planning. Starting with version 4.1, it is now the preferred user interface for EnergyScale functions on POWER6 servers and runs from within IBM Director 6.1. The way this works is that one can connect one instance of Active Energy Manager to many other platforms, including System i®, System p®, System z®, and System x®. There are client agents for each type of environment. Some new features as a result of this integration include the ability to configure events for power, temperature and CPU speed, power-save mode for POWER6 servers, extended power capping ranges, and command-line support. It’s also important to note that the IBM System Director Console for AIX is included as part of AIX 6.1 and all you need is a functioning Web browser in order for it to work.
AIX 6.1 – Workload Partitions (WPARs)
Virtualization unquestionably saves a lot of money on power and cooling. The more partitions you can carve on your existing systems, the better utilization you will have. WPAR, a new innovation of AIX 6.1, allow for operating systems virtualization, outside of traditional hypervisor-based logical partitioning. In essence, it allows you to carve up multiple workload environments inside of one host and OS kernel. So you can now have one single instance of AIX 6.1 and using WPARs you can have four different test environments working within that instance, appearing like separate operating system instances. Prior to WPARs, you would need to carve out separate logical partitions (LPARs) within the physical server.
This new feature of AIX 6.1 has the effect of reducing the amount of physical devices necessary to support your logical environments. Though one can conceivably get by with the bare minimum of physical devices for a logical partition, the WPAR is clearly the most efficient means that you can use when creating a new environment. This is purely an AIX innovation, unrelated to the POWER6. Furthermore, Live Application Mobility, another AIX 6.1 innovation, allows you to relocate running WPARs from one LPAR to another. IBM is the first vendor that allows for this feature without having to shut down the application or partition. This helps promote energy savings by allowing datacenter operators to move around workloads in such a way as to allow physical servers to sleep during non-peak periods. This feature is available on Power5™ servers and even Power4™ as well!
Consolidation and virtualization: Servers and datacenter
Unquestionably, server consolidation and virtualization go hand-in-hand. The bottom line is that server consolidation allow users to scale vertically — by consolidating to fewer more scaleable servers, from larger, more inefficient, and scattered traditional server farms. The way it can do this is through PowerVM, IBM’s Power virtualization product, which supports both AIX and Linux® operating systems. Server consolidation means doing the same or more with fewer hardware resources. Fewer hardware resources mean less money to spend on electrical power and cooling for physical devices. IBM has already released significant case studies around their POWER servers, proving how energy savings and POWER go hand-in-hand.
Pacific Gas and Electric consolidated 300 servers into six POWER servers and estimates are in for an 80% reduction in energy consumption (see Resources). Further, IBM’s Door Heat eXchanger water-cooling technology on POWER servers reduced heat by approximately 60%. PowerVM, through IBM’s hypervisor-based virtualization method, allows you to create as many as 10 logical partitions on one physical core. It also supports Symmetric Multi-Threading (SMT), which allows two separate instruction sets to run on the same physical processor — further increasing throughput by approximately 30%, which allows for the ability to function with less hardware.
There has been a trend in recent years away from distributed systems and toward the datacenter, particular with the rise of servers and datacenter consolidation. In the 1980s and early 1990s, a typical datacenter was usually very light, and there were distributed servers all over the place. When businesses realized this was not efficient — there were all sorts of issues from resiliency, to cost of maintenance, to disaster recovery, to systems not being backed up at all — the pendulum started to swing again. What was the effect of this?
The datacenter itself needed to be equipped to handle this change: power, cooling, and footprint. How much does power does a POWER server consume anyway? An IBM p575 weighs as much as a typical car providing 32 kW per hour, which is enough to heat most family homes. IBM has started to offer preconfigured modular datacenters (500 to 1000 square feet) that can actually be configured in less than four months. So how does datacenter consolidation help Green? The fewer datacenters and servers that you maintain, the lower the datacenter costs. Because the POWER architecture supports AIX and Linux, it further allows for many more consolidation opportunities. This is because it allows you to consolidate Linux operating systems around the POWER architecture and away from those commodity PCs that are never maintained properly. A new innovation around PowerVM is PowerVM Lx86, which allows you to run x86 Linux applications on POWER without needing a native compile. This feature really brings forth the ability to consolidate around the Power architecture in a huge way!
IBM: Big Green innovations
The IBM Big Green team applies science, modeling tools, and business expertise to address environmental management opportunities of all types. Among other areas, they are presently working on Advanced Water Management, Alternative Energy, and Computational modeling. Some recent highlights include:
* In mid-March (2009), IBM announced a new Public Sector Energy and Enviremental Diagnostic consulting server and strategic water information management solutions platform — two new initiatives to try to help the environment.
* IBM has just started the strategic carbon management service, a consulting arm that helps business formulate plans around smarter consumption of carbon. It does this by helping companies outline strategic goals around how to reduce energy use and carbon emissions, along with reduced cost. IBM’s goal is to bring emission savings to 50%. Rising energy prices are obviously a big factor. Lastly it has been shown that carbon, a major cause of greenhouse gas, is also a cause of climate change in terms of global warming.
Summary
For years, green computing has been a staple of IBM’s overall datacenter strategy. In recent years, through a myriad of innovations on the midrange, IBM has now brought green computing to Power Systems and AIX. The POWER6 and AIX 6.1 has been the impetus from which IBM has been able to provide many of the green computing initiatives through the past several years. IBM has also integrated technologies such as EnergyScale into their POWER architecture, which further has allowed for green opportunities. Datacenter and server consolidation projects have also been a driving factor towards green computing, as these projects easily leverage many of the benefits of green computing. These benefits include cost savings as well as environmental factors and regulatory drivers. IBM clearly has positioned themselves to continue to be on the forefront of innovation around green computing.
« IBM Director 6.1 – Management tool to reduce TCO | Home | Technical Posts – Contact Spectrum for further infromation »
Leave a Comment