/ Microsoft Right-Sizes its Private Cloud, Saves $3 Million and 9,000 Tons of Carbon Emissions
Overview
Country or Region:China
Industry:Software engineering
Customer Profile
Established in 2005, the Server and Tools Business division at Microsoft in China is dedicated to global and local development and innovation of core platform products and technologies.
Business Situation
The team faced inefficiencies with a decentralized server environment that supported 10 teams and more than 800 under-utilized physical servers.
Solution
The Research and Development Engineering Laboratoryimplemented a private cloud infrastructure built on Windows Server 2008 R2 with Hyper-V and Microsoft System Center products.
Benefits
- Increased server usage by 80 percent
- Simplified infrastructure maintenance
- Reduced hardware costs—U.S.$3,150,000
- Created a sustainable IT environment
Michael Wang, Director of Business Operations, Server and Tools Business, Microsoft Asia-Pacific Research and Development Group
In the Server and Tools Business division at Microsoft in China, the decentralized server infrastructure, which was used by 10 product teams, had grown to more than 800 physical servers—most of which were under-utilized—in just two years. The Server and Tools Business team decided to move to a private cloud infrastructure built on the Windows Server 2008 R2 operating system with Hyper-V and Microsoft System Center products. By the end of 2010, the team had more than 1,500 physical servers, 413 of which were hosts running 5,417 virtual machines. With its private cloud infrastructure, the team increased its server utilization by 80 percent, reduced the time that engineers spent on maintenance tasks by 70 percent, and saved U.S.$3,150,000. The team also created a green IT environment and saved 9,000 tons of carbon emissions and 14.85 million kilowatt hours of energy in one year.
Situation
At Microsoft, the Server and Tools Business in China began its journey in early 2005, with a vision of being a world-class development, innovation, and business center. The Server and Tools Businessin China is one of the core divisions in the Microsoft Asia-Pacific Research and Development Group and has more than 400 full-time employees in Shanghai and Beijing. The teams that are a part of the Server and Tools Business contribute to core platform products, including the Microsoft System Center family of products, Microsoft SQL Server data management software, Windows HPC Server, Windows Embedded Standard, and Windows Small Business Server. With its new mission to “Cloud-Optimize Every Business,” theteam is increasingly contributing to Windows Azure and Windows Azure–based services.
By 2007, the Server and Tools Business division in Chinahad accumulated more than 800 servers to support development and testing services for the rapidly expanding organization. Even though the team experienced explosive infrastructure growth in two years, a majority of the responsibility for setting up, configuring, and maintaining the servers fell on a small, three-person team of engineers in the Research and Development Engineering Laboratory.
Despite having more than 800 servers and facing continued growth, the organization’s utilization rates for the individual servers were low. The three-person team had to build server-based testing environments for a dozen individual product groups, each with specific development and testing objectives. This meant that the server infrastructure was decentralized, which resulted in inefficient resource management. “The root problem was that, instead of sharing server resources between product groups to maximize utilization, we deployed new servers for each new project, which effectively lowered utilization rates for individual servers to an average use-rate of 7 percent,” explains Michael Wang, Director of Business Operations, Server and Tools Business, Microsoft Asia-Pacific Research and Development Group.
Not only did the dispersed server infrastructure lead to under-utilized server resources, but it also put a productivity strain on the three-person team responsible for maintaining the infrastructure. The team spends a significant amount of time managing issues related to security, service pack upgrades, software updates, system restores, and data storage and backup. Managing the servers often meant repetitive, manual tasks that were tedious for the engineers. In addition, the Server and Tools Business at Microsoft China relied on multiple suppliers for server hardware, with many server configurations. In fact, any one product group might use up to 20 different server configurations using hardware from both Dell and HP. This meant that the engineering team did not have uniform management over the server hardware and would have to contact several hardware suppliers and coordinate between them. The process was not only time-consuming, but it also led to stresses on the laboratory and increases in the time it took to troubleshoot problems or deploy new services to support the business.
While the engineers faced the burden of managing more than 800 under-utilized, decentralized servers from a variety of manufacturers, the development and testing teams on different product teams faced inefficiencies with the server infrastructure. Because each product group had its own dedicated and independent servers, each with unique security settings and system access management controls, the groups could not effectively share troubleshooting knowledge between them. With the lack of uniform allocation for the servers, solving unexpected problems meant any one group often had to spend time troubleshooting and resolving the same issues that another group might have already resolved on their own servers. As a result, issues that might have taken only a few minutes to resolve with a standard, shared environment instead took hours to fix. This led to protracted service times that had a negative impact on internal customer satisfaction.
Solution
The Server and Tools Business team at Microsoft China needed to improve the infrastructure of the entire Research and Development Engineering Laboratory to mine the potential of its existing server resources, reduce the time it spent managing infrastructure and deploying services, and improve internal customer satisfaction. As a leader in developing enterprise-class cloud computing products, the Server and Tools Business team’s objective was clear: use the Microsoft private cloud platform as a foundation for a virtual server infrastructure.
“We used our own product development skills and best practices, combined with Microsoft products and technologies to create a four-step process for optimizing the server infrastructure for the entire laboratory,” says Wang. The goal was to steadily promote and then build a private cloud infrastructure without having an impact on the organization’s primary business objectives of product development and innovation.
Step 1: Server Consolidation and Virtualization
In 2007, the team first tackled server utilization and introduced a server virtualization platform based on Microsoft Virtual Server 2005 as a way of consolidating servers and optimizing server resource allocation. During the initial period, the laboratory consolidated 50 physical servers down to just four physical servers running 50 virtual machines. At the same time, the laboratory standardized its hardware-purchasing process and consolidated the number of suppliers it used for server hardware to only two. Each vendor uses the same specifications so that the Server and Tools Business team can more effectively manage a uniform hardware environment.
The laboratory and product groups subsequently decided to implement a virtual infrastructure on an even larger scale and, by the end of 2007, began using Hyper-V technology for its virtualization efforts. By the end of 2008, the teams had moved more than 70 percent of its core application servers, including database servers, web servers, and email servers, to the Windows Server 2008 R2 operating system with Hyper-V. It repurposed seven remaining physical servers to build a storage area network (SAN) service.
By the beginning of 2010, the server infrastructure had grown considerably from the 800-plus physical servers that the team used in 2007. However, the Server and Tools Business team no longer used a one-to-one ratio of physical servers to each server workload. The Research and Development Engineering Laboratory had more than 413 physical machines, which were used as virtual hosts to run 5,417 virtual machines.
Step 2: Automated Management and Resource Allocation
After the Server and Tools Business team consolidated its servers and virtualized a majority of its server infrastructure, it maximized its existing hardware investment by using the remaining server hardware as a resource pool. When a product team needs additional server resources, it can draw upon the server resources from that pool and, when the resources are no longer needed, they are re-pooled for another team to use.
To further maximize server use and to address the inefficiencies of decentralized infrastructure management, the team implemented Microsoft System Center Virtual Machine Manager 2008 R2. By using Virtual Machine Manager, engineers can manage the physical and virtual infrastructure components and centrally manage and allocate server resources to help further improve server utilization (Figure 1). Engineers can monitor virtual hard disks, virtual floppy disks, image mirrors, and virtual machine templates for all of the servers across all of the product groups from a single location. In addition, engineers can quickly deploy virtual machines to product teams with standard configurations and dynamically allocate additional server resources as needed to ensure an infrastructure that is the right size for utilization needs.
The team also uses Microsoft System Center Operations Manager 2007 to manage its server assets. Specifically, it monitors the state, health, and performance of is server infrastructure and gains deep insight into performance and utilization rates so that engineers can better optimize IT service delivery to the individual product teams.
Step 3: Self-Service Capabilities
With its largely virtualized server infrastructure and automated management capabilities in place, the Server and Tools Business team at Microsoft China had the core components of a private cloud infrastructure for its Research and Development Engineering Laboratory. Next, the team used the Windows PowerShell command-line interface software development kit to build a self-service portal that product teams can use to access System Center Virtual Machine Manager 2008 R2. Through the portal, the product teams can test development work through the common virtualization infrastructure, instead of building test scripts from scratch each time or relying on the three-person engineer team. In fact, they can check available resources and build a test environment in only four clicks.
By using the self-service portal, the development and testing engineers in each product team can request new virtual machines—either online or with a command line—and configure parameters for existing virtual machines according to their needs. They no longer have to rely on the over-burdened laboratory engineering team to configure a new physical server for each new request.
Step 4: Large-Volume Storage Allocation
The next step for the Server and Tools Business team was to address resources for the enormous volume of storage needs the team has. With a scalable private cloud infrastructure for server resources, the team wanted the same level of scalability for storage resources, including the ability to share storage resources across the entire Server and Tools Business division.
The team implemented Windows Storage Server 2008 R2, which is built on Windows Server 2008 R2, to build a SAN disk pool with a capacity of more than 100 terabytes. Product teams use the storage pool for the purposes of file-sharing and single-instance storage for storage merging and also for a SAN. By using Windows Storage Server 2008, the team can automatically discover and merge—or deduplicate—redundant files to optimize the storage space it uses.
Benefits
By building a private cloud infrastructure on Windows Server 2008 with Hyper-V and Microsoft System Center products for its Research and Development Engineering Laboratory, the Server and Tools Business team at Microsoft China optimized how it uses server resources. It moved from a decentralized server environment, where it not only under-utilized servers but also spent significant time managing dispersed servers across product teams, to a centralized, cost-efficient, energy-efficient, virtualized environment where many tasks are automated and resolved through self-service capabilities.
Increased Server Utilization
As a result of using a private cloud infrastructure, the Server and Tools Business team at Microsoft China has a right-sized infrastructure that is optimized to take full advantage of existing hardware, software, and storage resources. Whereas previously, servers ran on average at 7 percent utilization, they now run at 80 percent. “By using a private cloud infrastructure, we improved the utilization rates on our physical servers by more than 80 percent—without sacrificing server performance,” says Wang. “Now, we have to purchase new hardware only when utilization of physical resources reaches a specified limit so that every investment produces a better return on investment.”
In addition, by using a centralized storage resource pool, the team saved 4.5 terabytes of space on a single core server; by 2011, the team had saved a total of 15 terabytes of space by sharing storage resources across the division.
Simplified Infrastructure Management
Software development and testing personnel now use the Self-Service Portal to quickly access resources as needed, thus reducing human interference when an environment is set up and effectively improving the productivity of engineers. In addition, many of the previously manual tasks, such as provisioning servers, are now handled automatically, thanks to Virtual Machine Manager. For example, it now takes the Server and Tools Business team an average of only 15 minutes to set up a new development and testing environment, compared to an average of 2.5 hours previously. “With our private cloud infrastructure and automatic provisioning and other self-service capabilities, engineers spend 70 percent less time on infrastructure maintenance and management tasks,” says Wang. “That is 70 percent more time that they can spend focusing on developing and innovating on valuable solutions for our customers.”
Reduced Hardware Costs by More Than $3 Million
The Server and Tools Business team in China reduced costs by creating a private cloud infrastructure that does not require a one-to-one ratio of servers for each IT service, by further reducing the number of physical servers required through optimizing its server usage and by optimizing storage. For instance, to run the same infrastructure with an all-physical server environment, the Server and Tools Business team would need an additional 3,000 servers to the 413 physical servers it already has, at a cost of U.S.$1,000 per server—a savings of $3 million dollars.
For its storage, the team is using 150 terabytes of space. To otherwise support 150 terabytes of data in a physical environment, the team would spend $10,000 for every 15 terabytes of storage, or $150,000.
“Between server and storage hardware cost savings, building a private cloud infrastructure on Microsoft products and technologies has saved us $3,150,000,” says Wang. “That doesn’t even take into consideration that our infrastructure is still growing, which will result in continued savings.”
Created a Green IT Environment
By creating a private cloud infrastructure built on Microsoft products and technologies, the Research and Engineering Development Laboratory in China reduced its carbon-dioxide emissions by more than 9,000 tons in 2010 alone. In addition, the laboratory saved 14.85 million kilowatt hours of energy by significantly reducing the number of physical servers that it required to support its ever-expanding operations.
“The average household in China uses 2,500 kilowatt hours per year,” explains Wang. “This means that the energy savings achieved by our team were enough to power 5,941 homes in China for one year—a true testament of our commitment to pursuing green IT.”
Microsoft Cloud Power
Microsoft offers a complete set of cloud-based solutions to meet business needs, including solutions for advertising; communications (email, meetings); collaboration (document storage, sharing, workflow); business applications (customer resource management, business productivity); data storage and management; and infrastructure services. In addition, customers can take advantage of an entire ecosystem of solution providers and Microsoft partners.
For more information about Microsoft Cloud Power, go to: