Windows Server 2008 R2 Green Features
To the uninitiated, the so-called green features installed with Windows Server 2008 R2 might seem like nothing more than corporate charity aimed at environmentalists. While companies are all too eager to bask in the good guy status generated by incorporating energy saving measures and processes into their products, there is actually a significant business reason driving the inclusion of these features in products like Windows Server.
For decades, corporations have been building data centers and server rooms complete with raised floors and their own separate air conditioning systems to deal with the large amounts of heat generated by powerful servers, storage systems, and other components. These rooms were driven by necessity because overheating can cause the most stable of systems to fail. And, as processors became faster, and hard drives spun even more quickly, the amount of heat they generated grew by leaps and bounds.
Along the way, something else began to emerge, data centers with their own separate power supplies and generators. At first, these innovations were driven by the need for 24/7 server usage and a business necessity to achieve as near-zero downtime as possible. But, as data centers and server rooms grew, there emerged another necessity, the ability to get enough power to these massive computer centers without having to completely re-engineer the whole building’s power distribution system.
On the smallest of scales, the new power management capabilities built into Server 2008 R2 offer only small energy savings at best. A single server running the fastest processors, hard drives, and components won’t draw much more power than that old space heater the receptionist on the third floor keeps under the desk. However, like many things, when one stops to add all of the little differences up, the result ends up being a very big difference.
Recently, Google applied to U.S. regulators to become a national utility for the purposes of electricity generation and distribution, and more importantly, participation in the wholesale power markets.
Conspiracy theories aside, the move was largely driven by necessity because Google’s enormous datacenters require so much electricity that simply relying on the local utilities is becoming impractical.
While most companies may not have such dramatic power needs, server rooms and the numerous computers within them have become a major budget item, due not to typical maintenance and administration needs, but because of how much electricity they take to run. With the cost of electricity continuing to rise, and IT Departments looking to save money in their budgets, the power consumed by servers can no longer be ignored.
The typical server without power management features continues to draw between 50% and 70% of its maximum power usage even when completely idle. In other words, a server doing nothing uses at least half as much power as one running flat out. That means that every print server, domain controller, file server, Exchange server, etc., is sucking up plenty of power every minute of every day.
In terms of dollars, if servers were employees and electricity were their pay, it would be like paying every employee their full salary for the 40 hours per week they work, and then paying them half that rate for the remaining 128 hours. In other words, the amount of money spent to power servers when they are doing nothing exceeds the amount of money spent powering them while they are performing mission critical functions.
It’s no surprise then that customers both big and small are looking for ways to lower the power usage of their servers, not just to be good corporate citizens, but in order to save on the bottom line.
Server 2008 R2 comes with support for all of the latest energy saving standards, including updated support for Advanced Configuration and Power Interface (ACPI) in the form of a new processor power management (PPM) engine. When a server’s load is low, the processor is slowed down to consume less power.
Entire cores can be shutdown, a process called core parking. Likewise, hard drives can be spun-down to a low power state, and even the storage devices within SANs can be power managed.
To help businesses tell which products are capable of supporting these full power-management features, Microsoft has introduced the Enhanced Power Management Additional Qualifier (AQ) logo program. To use the logo, vendor’s hardware must support commands from the OS that reduce power during low utilization.
These are impressive technological improvements that demonstrate the commitment of the entire industry to reducing power consumption. However, to achieve significant power savings, it isn’t enough to just throttle back the processor when it isn’t busy.
Big power savings come from rethinking the entire IT datacenter architecture.
Rethinking Datacenters With “Green” In Mind
One of the first questions asked by non-IT executives when informed about the issue of server power consumption is why there are servers doing nothing in the first place. It is a fair question.
The days of slower WAN connections, and the increasing need for zero downtime led to a proliferation of “just in case” servers, or in more technically friendly-speak, over provisioning. Keeping server loads low ensured snappy response time. Having “extra” servers on hand in the form of fail-over systems, whether automated or manual, meant less down time. Running servers at 100% or even close to 100% was considered a sure-fire recipe for failure.
However, each of these goals also ensured that most servers were never fully utilized, creating an environment where servers were purchased, provisioned, and maintained based on how well 50% of the specs would meet the businesses current needs. (The issue of over-budgeting and “buying ahead” still exists, but is more of a corporate culture issue than a technology specific challenge.)
Fortunately, Microsoft’s continuing emphasis on developing better management and fault-tolerance within the Windows Server OS, also provides big opportunities for power reductions.
Server virtualization technology allows for more roles to be consolidated on a single server. Instead of one print server, one domain controller, one Exchange server, one SQL server, and so on, for each location, virtually all of those roles can be handled by a single server, without the fear of one service taking down all of the others should something go wrong, or increasing security complexity.
Server level hardware problems that cause a complete server failure become less common every year it seems. Even so, total fault tolerance in the form a fully redundant fail-over solution still only requires two servers instead of several.
Beyond Server Power Management
Enterprise level management tools included with Server 2008 or available separately can be used to see which servers are using the most, or the least power, and provide a starting point for where to being consolidation or virtualization. As an added bonus, these same management features also improve monitoring, reporting, and alerting, providing not just potential power savings, but increased uptime and easier management as well.
While most IT Departments are currently focused on the power consumption of their servers and data centers, awareness is arising as well about the power wasted by workstations which often operate at full power even less often than servers do.
Management tools within Server 2008 and Windows 7 allow for IT to begin managing this cost as well. Group policies, for example, can be set to move workstations into standby mode or to power off completely after their backup or other “off-hours” tasks have been completed.
The opportunities for power-savings are enormous. Savvy IT Departments should consider having data center electricity moved into their own budget lines allowing them to both “get credit” for significant expense savings, as well as giving them the justification to migrate to important new technologies like Server 2008 R2 and Windows 7 in the form of large operating expense reductions.