Export (0) Print
Expand All

Microsoft IT Leverages its Compute Service to Virtualize SharePoint 2010

Article

Published: June 2011

Microsoft IT piloted the deployment of a virtualized Microsoft SharePoint 2010 environment using the Compute Service, which provides high density compute servers and virtual machines (VMs) to the business. The SharePoint team saw the collaboration as a way to reduce operational costs and complexity, and the Compute Service team viewed it as an opportunity to significantly enhance its infrastructure capabilities. In addition to driving down costs, the partnership helped mature the Compute Service, because it allowed the team to identify and address a business gap by upgrading its physical infrastructure.

Download

Intended Audience

Download Article, 289KB

Intended for IT Professionals and IT Managers, this article provides insight into the technical and process-related considerations and benefits associated with deploying a virtualized SharePoint 2010 environment using the Compute Service.

Introduction

Microsoft IT (MSIT) manages one of the largest implementations of Microsoft SharePoint® in the world. MSIT SharePoint farms serve North America, South America, Europe, the Middle East, Africa (EMEA), and Asia Pacific (APAC). The implementation hosts over 36 terabytes of content in 250,000 site collections. SharePoint provides the foundation of most collaboration at Microsoft.

The MSIT Compute Service fulfills high computing needs by allocating, deploying and managing high density compute servers and Virtual Machines (VMs) to the business. As part of the MSIT commitment to virtualize 50% of its infrastructure by July 2012, the Compute Service team currently manages over 11,000 VMs and 2,100 VM hosts within Microsoft IT-managed data centers.

Because SharePoint is so crucial in facilitating collaboration at Microsoft, the MSIT SharePoint team takes cautious and measured steps when modifying the service in any way. Service changes are carefully weighed against any potential impact to users, and to the business. The same caution applied when utilizing the Compute Service to introduce virtualization into the SharePoint environment.

Minimizing Impact

To minimize the impact of virtualization activity on users and the critical business applications that rely on SharePoint, the SharePoint and Compute Service teams chose to introduce virtualization on a "pilot in production" server farm. Although the "pilot in production" SharePoint implementation is a dedicated test environment, it serves about 130,000 unique Microsoft users. The farm itself hosts approximately five terabytes of data, in 5,800 site collections and 28,520 sub sites. While not a true production implementation, the server farm provides the significant capacity and usage stresses required to test an enterprise environment.

Meeting Infrastructure Requirements

SharePoint requires a very robust infrastructure. MSIT had to ensure that the Compute Service infrastructure met the minimum requirements to virtualize the SharePoint environment. Virtual machine capabilities are heavily influenced by underlying physical hardware. Prior to the pilot, the SharePoint servers had not been virtualized. This was due to the large number of processors and memory required.

When the SharePoint team began to plan for its transition to a virtualized environment, the Compute Service infrastructure was unable to support it. The SharePoint team provided their server requirements for IIS, Index, and Application servers.

The Compute Service then upgraded their infrastructure, offering a comparable virtual machine configuration, as illustrated in Table 1.

Machine Type

CPU

Memory

Operating System

NIC Speed

Physical Server

Intel Xeon E5410 @ 2.33 GHz

(2 processors — 8 cores)

32 GB

Windows Server 2008 R2 x64

1 GB

per second

Virtual Machine

Intel Xeon L5530 @ 2.4 GHz

(4 cores)

16 GB

Windows Server 2008 R2 x64

10 GB

per second

Table 1. Comparison of physical server and virtual machine requirements for SharePoint 2010

By significantly enhancing its underlying infrastructure, the Compute Service moved into a position where it could support the significant availability and consistency requirements of the business-critical SharePoint service. By 2010, the Compute Service was ready to host the virtual SharePoint environment, and within a couple of months the "pilot in production" farm was online.

Phased Deployment

The SharePoint virtualization deployment was directly proportional to the customer workload and the number of connections to the server farm. It took place in phases. For every two virtual machines deployed, one physical server was decommissioned. The overall deployment process was similar to a physical server deployment. Once the logical servers were procured, the Compute Service team managed the deployment in phases per the SharePoint team's requirements.

Figure 1. Timeline of phased virtualization deployment

Figure 1. Timeline of phased virtualization deployment

Rolling out Applications

When deploying some of the applications on SharePoint, including PowerPivot for SharePoint, Microsoft Project 2010, Microsoft Office Web Apps, PerformancePoint Services for SharePoint, and Microsoft Visio® 2010, it was initially difficult to gauge the number of users that would access an application at any given time. Resource utilization also needed to be factored in, as the applications tended to use more resources than simply using a web browser to view SharePoint pages.

To mitigate some of the application rollout challenges, MSIT completed these steps:

  • Isolated lab testing was performed on applications

  • Applications were deployed to a pre-production environment

  • Applications were deployed to production virtual machines

Based on CPU, memory, disk, and network performance, the SharePoint team was able to determine whether they needed to add more servers into the rotation, or place the application services on different application servers. For example:

  • If the CPU, memory were operating at greater than 70%, MSIT would add another server into the rotation.

  • If MSIT observed network queuing, more servers would be added into the rotation, or servers would be moved around different VM hosts in order to get more network throughput.

  • If disks were queuing requests, or had low IO per second, MSIT would engage the Shared Storage team to troubleshoot slow disk performance.

As shown in Figure 2, the SharePoint team determined the appropriate and efficient blend of services, where customers were not impacted by underperforming servers or virtual machines.

Figure 2. SharePoint VM Topology

Figure 2. SharePoint VM Topology

MSIT ensured that all key services were running on at least two VMs, particularly the more intensive services, such as Office Web Apps, Project 2010, and PowerPivot for SharePoint. MSIT followed a best practice of deploying those services either onto separate VM hosts, or onto a host with another service that is not as resource-intensive. The Compute Service was able to accommodate physical redundancy necessities that were provided within the deployment requirements.

Improving Daily Operations and Management

One of the benefits realized from virtualizing SharePoint was simplified daily operations and management. The SharePoint service teams were initially challenged because significant service responsibilities transitioned to the Compute Service team. Traditionally, the SharePoint service teams had controlled and managed ongoing server operations and troubleshooting.

The SharePoint team learned to adapt to a new support and escalation model for their VMs and associated host servers. In the new model, the SharePoint team disassociated itself from server maintenance and troubleshooting. Instead, a separate support team receives automated server health and welfare alerts, which allows for proactive management of incidents related to hosts and VMs.

In parallel with the SharePoint virtualization process, other systems supporting the Compute Service's request management and service management infrastructure were being upgraded to enhance operational efficiency. This introduced some temporary challenges and some minor delays for the SharePoint team. However, this was not surprising, due to a new service engagement being piloted with another team that was itself working through new processes and systems.

The SharePoint team was able to pilot new Compute processes, and to provide meaningful feedback to the service operations team. Ultimately, the Compute team was able to incorporate key changes into its own service management system.

Initial Results

The initial results of the "pilot in production" deployment have been very favorable. Availability has been high, averaging around 99.9%, and the service is stable with an average latency of less than .03 seconds for most page loads. The latencies were consistent, due to the reliability of the Compute Service, as well as the consistency of the deployed virtual machines.

High Availability

The Compute Service provides high availability through Microsoft System Center Virtual Machine Manager live migration support. During VM host maintenance, such as during security patching, live migration support allows the SharePoint service to remain online, subsequently meeting the necessary availability requirements.

Capacity and Management

It is much easier to set up and add capacity to a new VM than it is to procure and install a new server in a data center. Overall hardware and hosting costs are also reduced, as there is an 8:1 physical to VM ratio.

Because the Compute Service manages the physical infrastructure of the hosting environment in its entirety, VM owners don't need to concern themselves about whether they are managing a VM or a physical server.

Cost Savings

While it is still too early in the process to provide actual numbers around cost savings for this specific virtualization effort, the average monthly savings per VM within MSIT data centers is $350 USD. This accounts for 3.89M$ in MSIT monthly cost avoidance at the current virtualization rate of 47.40%. All indications are that the combination of reduced hardware, reduced power consumption, and service management simplification will lower the cost to implement and manage a virtual SharePoint system.

Conclusion

Virtualization is an important strategy for reducing space and power consumption in Microsoft IT data centers, and for ensuring appropriate server utilization. The success of this pilot proved the feasibility of deploying virtualization as part of the SharePoint utility strategy

In addition to the anticipated cost savings realized by the pilot, the partnership enabled maturation of the Compute Service, because it allowed the team to identify a business gap and work to bridge it by upgrading its physical infrastructure.

Because of the successful virtualization of the "pilot in production" server farm, MSIT has begun moving forward with virtualizing additional farms in North America, EMEA and APAC regions.

For More Information

For more information about Microsoft products or services, call the Microsoft Sales Information Center at (800) 426-9400. In Canada, call the Microsoft Canada Order Centre at (800) 933-4750. Outside the 50 United States and Canada, please contact your local Microsoft subsidiary. To access information via the World Wide Web, go to:

http://www.microsoft.com

http://www.microsoft.com/technet/itshowcase

© 2011 Microsoft Corporation. All rights reserved.

This document is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. Microsoft, SharePoint, Visio, Windows, and Windows Server are either registered trademarks or trademarks of Microsoft Corporation in the United States and/or other countries. The names of actual companies and products mentioned herein may be the trademarks of their respective owners.

Was this page helpful?
(1500 characters remaining)
Thank you for your feedback
Show:
© 2014 Microsoft