SharePoint 2013 Microsoft WAN Case Studies

  • 876 views
Uploaded on

 

More in: Technology
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Be the first to comment
No Downloads

Views

Total Views
876
On Slideshare
0
From Embeds
0
Number of Embeds
0

Actions

Shares
Downloads
23
Comments
0
Likes
1

Embeds 0

No embeds

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
    No notes for slide

Transcript

  • 1. SharePoint 2013 WAN case studies: United Airlines, Teck Resources, andFabrikamSummary: Learn how three global companies designed their SharePoint architectures to meet theirrequirements.Applies to:In this article: United Airlines—Central deployment, single datacenter Fabrikam—Central deployment, multiple datacenter Teck Resources—Geographically dispersed deploymentAttributionsUnited Airlines, Fabrikam, and Teck Resources worked with Microsoft in the prerelease program forSharePoint Server 2013. The IT teams for these companies were led by the following individuals.Denise Wilson — Senior Manager of Enterprise Collaboration Services, United AirlinesMahmoodJaffer — IT Specialist and SharePoint Architect for Teck Resources, Softlanding(www.softlanding.ca)Fabrikam is a fictionally named company representing a large world-wide manufacturing company. While theFabrikam IT team remains anonymous for this case study, the team was very active in providing valuablefeedback to Microsoft.Microsoft Senior Consultant, Faizan Khan, worked with these companies to evaluate WAN performance andrecommend appropriate architectures using SharePoint 2013.―Getting a chance to work on the latest version of SharePoint with three global customers with uniquescenarios was rewarding. I enjoyed tackling the previous SharePoint 2007 and 2010 paradigms related toarchitecture and WAN connected users head on with all of the new SharePoint 2013 capabilities. It was greatto see how SharePoint 2013‘s higher scalability and optimizations for connectivity over the WAN allowedeach customer to reduce their server footprint while providing an excellent social collaboration experiencefor their own customers. When you start planning your SharePoint 2013 architecture I highly recommend
  • 2. taking a look at the new opportunities to consolidate your infrastructure as these customers have done.‖ –Faizan Khan United Airlines—Central deployment, single datacenterUnited Airlines is one of the world‘s leading airlines serving 379 global airports on six continents with over85,000 employees. United Airlines participated in the SharePoint Server 2013 Rapid Deployment Programand the Windows 8 and Windows Server 2012 Technology Adoption Programs.Denise Wilson, Senior Manager of Enterprise Collaboration Services team at United Airlines, worked withMicrosoft on several projects with SharePoint Server 2013. Wilson worked with Microsoft Consulting Servicesto architect the new SharePoint Server 2013 environment and to provide feedback about the airline‘sexperience using SharePoint Server 2013 in its WAN environment. Wilson also worked closely with theSharePoint product team to develop and pilot a new procurement tool that is based on taxonomy and newsearch features in SharePoint Server 2013.United Airlines‘ employees rely on SharePoint sites to complete job-related tasks at every company site,including the 379 global airports. The route map for the airline shows the geographic environment. Thewhole environment is served by a centralized SharePoint farm.The environment includes high latency connections and other challenges that relate to performance. Wilson‘steam addressed these challenges in many ways. The following table summarizes some performanceoptimizations that made a positive difference for the company. The third column of the table includesadditional options that other organizations with similar challenges can consider. Additional options that Challenges United optimizations other organizations can considerHigh utilization, low- Quality of service (QoS), browser compressionbandwidth connections at WAN accelerators (Internet Explorer)international locationsPerformance at airport SQL Server Master Datakiosks QoS, browser compression, local caching Services (MDS), network interface card optimizationPrioritizing customer kiosk
  • 3. performance over back officeperformance of SharePointtraffic Browser compression, Internet InformationIntranet SharePoint Services (IIS) File Compression (static and Branch Cacheperformance dynamic), binary large object (BLOB) Caching, Output Caching, Object Caching Do not enable auto-create Memory tuning, reducing databases, statisticsSQL Server performance reorganizing indexes, and rebalancing theoptimization virtual log files Set max degree of parallelism (MAXDOP) to 1The last row of the table, SQL Server performance optimization, highlights the importance of payingattention to the overall health and performance of a SharePoint farm. While these are not WAN-specificoptimizations, the United Airlines team reported noticeably faster SharePoint performance across theenvironment after optimizing compression and caching.United Airlines operates its SharePoint environment in a Chicago-based datacenter. The disaster recoveryenvironment is hosted in its Houston datacenter. Currently the company uses log-shipping between farms.Failover between farms is scripted and can be achieved in about 15 minutes. The following diagram showsthe disaster recovery environment.
  • 4. ―There are a number of options for supporting high availability and failover,‖ states Khan. ―This example, logshipping between farms across the WAN, is fairly standard. United Airlines has tiered their services anddatabases. Some databases have a high or medium SLA and are log-shipped. Others, such as contentdatabases, have a lower SLA and are backed up and then restored. While log-shipping is a common way tobuild a failover solution, database mirroring can also be used, as well as high availability groups with SQLServer 2012.‖The overall performance in SharePoint Server 2013 is optimized for building and operating larger farms.Even with the high number of users, large volume of content, and extensive use of the feature set, theMicrosoft Consulting team and the SharePoint product team feel comfortable recommending a single, largefarm for United Airlines.The farm featured in the Enterprise-scale model resembles the farm designed by Khan for United Airlines.This model provides guidance for patching and updating farms of this size.
  • 5. Zoom into the model in full detail with Zoom.it from MicrosoftVisio versionPDF version Fabrikam—Central deployment, multiple datacenterFabrikam is a fictional company that represents a large world-wide manufacturing company that participatedin the SharePoint Server 2013 prerelease program. Fabrikam has several-hundred-thousand employees andmany content farms that were migrated from the previous version.A priority for Fabrikam is to take advantage of the new social features in SharePoint Server 2013, such asoffering My Sites to all employees. With the large number of employees, much consideration went into thearchitecture for My Sites. Fabrikam weighed the benefits and risks of deploying one My Site farm versusmultiple My Site farms. The following table summarizes the advantages and disadvantages for botharchitectures. Architecture 1 — One large My Architecture 2 — Two or three My Site farms Site farm and multiple content and multiple content farms farms Two or more My Site farms are deployed and employees are spread across the My Site farms. Each My Site farm consumes a dedicated User Profile service application. The My Site farms and All employees are served by one My the User Profile service applications can be spread Site farm. The My Site farm and the across datacenters if the User Profile service content farms are located in the same application that is associated with a My Site farm isDescription datacenter and consume the User located in the same datacenter. Content farms can Profile service from a dedicated be spread across datacenters. However, each services farm which is also located in content farm must consume a User Profile service the same datacenter. application that is located in the same datacenter. The User Profile Replication Engine is used to synchronize profile data across the User Profile service applications. Single User Profile service to manage. My Site farms are smaller. Single My Site farm to manage. The environment can be distributed across multipleAdvantages datacenters. Less complexity—no need to use the User Profile Replication Engine. Users can be assigned to My Site farms based on location, which provides greater opportunities for With a single User Profile service
  • 6. application, all social features are maintenance windows. expected to work as expected across farms. More farms to maintain. Setting up and managing the User Profile Concern about outgrowing a single Replication Engine is complex. My Site farm.Disadvantages With multiple User Profile service applications, With a world-wide user base, its some social features might not work as expected more challenging to schedule across farms. For more information, see maintenance windows. "Deployment issues with multiple User Profile service applications" in Plan for My Sites in SharePoint Server 2013.An advantage of the performance improvements in SharePoint Server 2013 is the ability to combine contentand users into a smaller number of large farms. These improvements helped Fabrikam to feel comfortablewith a single My Site farm (Architecture 1 in the previous table). This architecture avoids the complexity ofmanaging multiple User Profile service applications. However, with a single User Profile service application,all of Fabrikam‘s content farms must be located in the same datacenter. The following diagram showsFabrikam‘s environment.In the diagram, a separate datacenter hosts Fabrikam‘s search farm and a dedicated search team managesthe farm.Khan explains the design challenge: ―Because the social capabilities of SharePoint Server 2013 areextremely important to Fabrikam, we engaged in a number of detailed conversations on the best approachto implementing a My Sites farm. Should we implement one large farm or fewer smaller farms? There arepros and cons with either choice including manageability, availability, and functionality. There are also
  • 7. concerns with scalability both on the application and storage layers. With new guidance from the SharePointproduct group, we decided the best approach is to implement a single farm. This approach results in thebest social experience while remaining within the scalability guidelines of SharePoint Server 2013. Due tothe large user base, Fabrikam highlights the ability of SharePoint Server 2013 to scale while minimizing thenumber of farms required.‖For more information about how to design architectures that have multiple farms and multiple datacenters,see the following model: Multi-farm architectures with SharePoint Server 2013.Zoom into the model in full detail with Zoom.it from MicrosoftVisio versionPDF versionA goal of working with Fabrikam in the prerelease program was to understand how SharePoint Server 2013performs over a variety of WAN connections. Fabrikam used Visual Studio to script a load test that is madeup of many unit tests and then ran the load test from multiple geographic locations. For more informationabout how to create a similar test, see SharePoint 2013 WAN testing with Visual Studio 2012 walkthrough.In this first set of results, two users in the Fabrikam Shanghai, China, office ran the load test against theSharePoint sites in the Texas, USA, datacenter. Latency is about 190 milliseconds roundtrip. The upload,download, and Outlook Web Access (OWA) tests were conducted with a 1 mb file.
  • 8. The test results show that performance is good, especially for the social tasks.The next set of results shows performance for the same load test across a larger set of geographic locationswhere Fabrikam employees work. The servers running SharePoint Server 2013 are located in Texas, USA.Even though there are varying degrees of latency, performance is good for users across the globe. TheFabrikam test results provide an example of systematic WAN testing by using a load test that is comprisedof a variety of SharePoint tasks that are important to the company.Fabrikam is an example of a world-wide company that is succeeding with a central datacenter model,instead of deploying SharePoint Server 2013 to multiple regions across the world. If you are considering amove from a central datacenter model to multiple SharePoint sites in different geographical regions, makesure that you conduct WAN testing to see whether it is really necessary.
  • 9. Teck Resources—geographically dispersed deploymentTeck Resources is a mining company with many remote sites throughout the world. Many sites areconnected to the Teck Resources WAN by high latency and intermittent satellite connections.MahmoodJaffer, IT Specialist and SharePoint Architect with Microsoft Partner Softlanding, worked with TeckResources to design and support their global SharePoint architecture.The description of the Teck Resources architecture includes several illustrations that might be difficult toread within this article. These illustrations are also available in SharePoint 2013 WAN testing with VisualStudio 2012 walkthrough.Teck Resources uses two datacenters in Calgary, Canada, and Santiago, Chile. The following map shows thetwo datacenters plus the 10 remote sites.Even though some remote sites are located on the same continent as a datacenter, these sites rely on lowbandwidth connections or satellite connections that are intermittent. For sites shown on the map, WANperformance prevents or discourages users from uploading content to a SharePoint site at one of the twodatacenters, or a site located at a nearby site.Teck Resources weighed the costs and benefits of deploying many regional SharePoint farms compared tonot using SharePoint sites. For the mining company, the benefit of using SharePoint Server 2013 features tomanage content outweighs the administrative costs of deploying many farms to support the distributedenvironment. In particular, the company relies on the following capabilities: Managed metadata—a standard tagging scheme that is applied to content so that assets can be discovered and managed appropriately over time.
  • 10. Enterprise-wide search—the ability to discover content assets that meet specific criteria, which metadata defines, and to know the mining sites that these assets are associated with.Integrating the environment to support these features requires coordinating service applications at all farms.The User Profile Service and My Sites are deployed to each farm. The User Profile Replication Engine is usedto synchronize profiles between the two datacenters, as shown in the following illustration.Because the environment is distributed among more than three farms, there is no attempt to synchronizeprofiles across all farms, that is, synchronize each farm with all other farms. Instead, each remote farmsynchronizes with only one of the two datacenters, as shown in the following illustration.
  • 11. With this highly distributed environment, Teck Resources is willing to accept that there might be somemanual cleanup work with profiles as described in Global architectures for SharePoint Server 2013.Search is an important component of the enterprise content management strategy of Teck Resources. Thecompany has to find assets that meet specific criteria and to determine the mining sites with which assetsare associated. Currently much of the historic data that belongs to Teck Resources resides on file shares. Itwill take some time for the company to move the content to a SharePoint environment. Some assets mightremain on file shares indefinitely.At the remote sites, search is deployed to each local farm and crawl rules are set up to also crawl local fileshares. The following diagram shows the configuration that is replicated for all remote sites.
  • 12. Enterprise-wide search is provided at the company‘s two datacenters in Calgary and Santiago. Search atthese datacenters is configured to use either federated search or remote result sources to display resultsthat are associated with each remote site. Remote result sources are used for the sites that have reliableWAN connections. Federated search is configured for sites with intermittent satellite connections. Federatedsearch is configured to show results in a separate Web Part which provides a visual indication of whether theremote site is connected or disconnected at the time of the search.The following diagram shows how enterprise-wide search is architected to include results from all remotesites.The relationship of the two datacenters to each other is not included in the illustration. Each datacenter usesremote result sources to include the other datacenter in search results.Finally, the Managed Metadata service application is integrated by deploying it to the central farm in Calgaryand then sharing this service application to all other farms in the environment. The Managed Metadataservice application is one of the few service applications that can be shared across farms that are separatedby WAN links.
  • 13. One issue that Teck Resources reports with this configuration, however, is that the Managed Metadata fieldsand tagging are not available to remote farms when a satellite connection is disconnected.Khan summarizes the design challenges for Teck Resources: ―Because Teck Resources supports thecollaboration needs of occasionally connected offices or offices with extreme latency due to the use ofsatellite links, we needed to build an architecture that was flexible while minimizing the complexity as muchas possible. The WAN improvements helped Teck Resources consolidate some of the remote farms, but notall. The result is two central sites that act as the social hub for a number of smaller farms. This provides asomewhat cohesive social experience while supporting the goal of providing remote office users with accessto SharePoint sites even when disconnected from the ‗hub‘.‖See alsoGlobal architectures for SharePoint Server 2013WAN performance and testing for SharePoint 2013