How to Plan Your Server-Consolidation Strategy with System Center
Frank Koch and Michael Walsh
The best way to save energy is to avoid unnecessary energy consumption. While this principle of Green IT may seem obvious, a recent study by the Uptime Institute suggests that many IT departments fail to implement solutions that adhere to this principle. In fact, the study found that 10% to 30% of servers in the data center consume energy without serving a function.
Microsoft System Center, coupled with virtualization software, can help you avoid unnecessary energy consumption and reduce IT costs through physical server consolidation.
Why Consolidation Matters
Say there is a business group that requires a new application or service. Until recently, many IT professionals would simply install a new physical server to handle the task without first checking to see if the application could run on an existing box. Why? Because the IT pro learned through experience that it was far easier and cheaper to procure a new physical server than to reprovision an existing server to handle multiple functions.
But today, energy costs for operating a server are often higher than the acquisition costs of the physical hardware, and so this reasoning no longer holds true. Now the IT pro requires solutions that consolidate physical services in order to improve efficiency, reduce wasted energy, and cut costs. And this is where Microsoft System Center comes into play—it offers a variety of tools for implementing consolidation strategies for both homogeneous and heterogeneous workloads.
Nowadays, consolidation of a homogeneous workload is very easy. Hardware can handle much greater load and user access than in the past. With 64-bit, multi-core processor technologies; Gigabit network cards; and vastly more RAM and disk space than in the past, a single server can often handle a load that once required several servers. This is true especially for file, print, Web, and terminal servers.
However, as with many technology solutions, the solution itself creates a new problem: the new higher load can exponentially increase the risks for system failure. Therefore, you need a complex system-monitoring solution capable of providing much deeper information than a simple system ping. Microsoft System Center Operations Manager (OpsMgr) provides a robust system-monitoring solution, supporting Windows and Windows applications as well as cross-platform monitoring.
It allows you to work with abstract IT services (such as messaging) instead of single systems (such as one dedicated Exchange Server). This enables you to take full advantage of consolidation efforts.
In addition, OpsMgr provides both service and performance monitoring. This lets you check if the network is saturated, whether the disk subsystem cannot handle the load, if the memory is getting low, whether the processor is overloaded, and so on. All this data is transferred into a data warehouse for long-term archiving to allow baseline creation and comparison of system load over time.
IT departments, therefore, can implement easier and safer consolidations of homogeneous workloads with the assistance of Microsoft System Center.
The consolidation of heterogeneous workloads is far more complex. The dependencies between applications are harder to predict, which is why people have often been afraid to attempt heterogeneous consolidation.
Advancements in virtualization software, though, have provided a technological solution that makes heterogeneous consolidation simpler—enabling multiple server instances to run on the same hardware simultaneously without affecting one another. Such software enables IT professionals to easily consolidate several machines running at a low utilization rate between 5% and 25% into a single physical box, increasing the overall utilization of the box by as much as 80% or higher. Given that several independent studies show that most servers today are running at a utilization rate of less than 20%, this presents an opportunity to dramatically reduce the number of physical servers needed per organization.
However, virtualization software alone does not provide a comprehensive heterogeneous consolidation solution. This is because virtualization software only solves one of the two obstacles to homogeneous consolidation—the ease of implementation. The safe execution of homogeneous consolidation via virtualization is equally important. Otherwise, consolidation will simply lead to more systems-maintenance issues and will negatively impact the availability of servers and the integrity of data. Therefore, you need a powerful systems-management solution.
Figure 1 Converting a physical machine into a virtual machine
Consolidating with System Center
The first step to consolidation is to find out which servers can and should be virtualized.
Ideal candidates will have complementary low utilization levels. So you need to plan your implementation strategy based upon server performance over time.
Performance monitoring with Microsoft System Center can help you visualize and analyze each server's performance over time, collecting data with unique performance counters per system or system group. With this data, you can easily generate predefined reports for virtualization candidates based on common criteria, such as disk, processor, and memory usage. This offers insight into which physical servers possess low utilization rates based upon patterns of use and activity. This is the key data you'll need for planning an optimal consolidation strategy.
Figure 2 Customizing host ratings for a virtual machine
The next step is to convert the servers into virtual machines and aggregate several of them on the same physical box.
The conversion of servers is simplified with the use of Microsoft System Center Virtual Machine Manager 2008 (VMM). To convert a physical machine into a virtual machine with VMM, all an IT Professional has to do is start a wizard and pick the right physical box. VMM will check the machine for any potential issues that might occur during the process and then convert the machine on the fly while the system is running, as shown in Figure 1. (This feature is available for current versions of Windows Server.) The machine is then available in the standardized Virtual Hard Disk (VHD) format and can be placed on any virtualization hosts, such as Hyper-V or Microsoft Virtual Server 2005 R2 machines.
While the physical to virtual conversion is nothing spectacular these days, the actual placement of the virtual system on the ideal host is a unique feature of VMM. The PRO (Performance and Resource Optimization) feature is utilized here. By combining the knowledge of VMM with OpsMgr, the ideal host for virtual machine combination is selected, and the decision is dynamically reviewed in case the load scenarios are changing. (You can customize how ratings are assigned, as shown in Figure 2.)
PRO analysis will select the appropriate combination of guest systems to avoid processor, disk, or memory overload of the host systems. This helps to avoid that key concern that heterogeneous consolidation will result in the combination of several memory-intensive applications on the same host or guests, causing heavy disk I/O or processor needs.
With OpsMgr and VMM, Microsoft System Center can help you achieve a combination of complementary utilization rates that will add up to an optimal load on components—without risking one bottleneck bringing everything down. OpsMgr provides service and performance monitoring of physical boxes as well as your virtual servers, regardless of whether they are running Windows Server or another supported OS, including VMware ESX.
Meanwhile, VMM makes management of large virtualization farms (with hundreds of hosts that may even be distributed over several locations) as straightforward as managing a single box. And again, this includes non-Microsoft virtualization technologies, such as VMware ESX.
In the end, you can consolidate server farms down to the minimum (ideal) number of physical hosts your environment needs. And this will reduce the amount of energy being consumed by your infrastructure.
Frank Koch is an Infrastructure Architect Evangelist at Microsoft. He resides in Germany.