Geographically distributed deployments of SharePoint 2010

SharePoint 2010 brings a large amount of improvements for geographically distributed SharePoint installations.

SharePoint deployments can be categorized in the following segments:

• Uni-centric deployments
  SharePoint is installed on a single data centre with a geographically dispersed user base.
• Multi-centric deployments
  SharePoint is installed on multiple data centres across regions/continents and clustered user bases close to the data centres.
• Hybrid deployments
  SharePoint is installed on multiple data centres with a geographically dispersed user base.

Globally dispersed users can be categorized into the following segments:

• Local/LAN users
  Users with high bandwidth, 1mb connections and up and low latency at less than 50ms per roundtrip
• Continental latency
  Users with high bandwidth (ADSL, cable) but higher latency at up to 125ms per roundtrip
• Global users
  Users with high bandwidth but with even higher latency at up to 300ms per roundtrip
• Low bandwidth users
  Users in developing countries, satellite links, mobile connections. These users would have low bandwidth and very high latency – too disperse to estimate

Out of these figures we can estimate WAN performance. These figures show degradation in time for the first page to load (PLT1) over different segments assuming that LAN users are the starting point at X seconds:

Segment	PLT1	Example
LAN	X	3 seconds
Continental	2X-4X	6 to 12 seconds
Global	4X-8X	12 to 24 seconds
Low band	??	Very hard to estimate

 

SharePoint 2007 deployments

The vast majority of SharePoint 2007 deployments used the uni-centric deployment as network related issues were too great to overcome. This is due to the nature of a farm, where all servers communicate with each other as a “single entity”. The farm has a “shared services provider” which is a site on the farm that contains services that are shared between web applications on the farm. These include searching and user profile storage and management.

SharePoint 2007 can be installed over separate databases during a global deployment so that team/department site databases can reside closer to its intended user base, but the shared services and farm configuration database must still be shared. This causes a lot of communication over global connections and keeping these at high speed and low latency is both costly and difficult.

New capabilities in SharePoint 2010

• Cross-farm/multiple instance Shared Services
  The “shared service provider” is not obsolete. Services are provisioned per web application and can be hosted either on the same farm or on a remote or “master” farm. The services can also be configured in multiple instances (for simultaneous access over load balancing) or customised per web application. Remote shared services, i.e. published services, also involve caching between farm instances for performance optimization. The most common shared services are user profile storage, managed meta data services (farm/cross farm shared content types and taxonomy), search and BDC/BCS connectivity services.
• Uninterrupted log shipping
  Creates read-only farms by using SQL log shipping of the entire farm. This can be very useful for creating local copies of content for users on remote continents. Log shipping sends a diffs (differential data packets) of the data and is therefore very fast. As an added bonus, you can use the read-only farm as a disaster recovery platform where you can turn off the read-write farm and make the read only farm the master read-write farm.
• FSSHTTP
  This catchy acronym stands for “File Synchronization via SOAP over HTTP” and works by reading/writing diffs between the SharePoint server and Office 2010. When reading a file, a checksum is sent to the server. If there are any changes, then only the change diff is downloaded to the end user. This will drastically reduce network usage when working with large documents and/or large numbers of documents.
• ODC
  Another acronym that stands for “Office Document Cache”. This application lives on the client machine and manages diff replication of documents between the client and the server. Normally, when saving a document, the Office application will wait for the document to finish saving. With ODC, the document is saved to the “save queue” and is synchronized in the background. This also allows for offline access of already opened documents as well as "offline saving”. Multi-master merges are solved by the ODC engine.
• SharePoint Workspace
  This is the new version of “Groove”, an application that looks like any other Office application but contains an explorer-like view of the SharePoint data. You can synchronize documents and lists for offline viewing and editing.
• Office Web Applications
  Documents can be viewed/previewed and edited directly in the browser using client script and AJAX. Not only does this mean that users do not need the full Office client on their local machine but it uses much less bandwidth.
• Mobile Access View
  The mobile version of SharePoint has been vastly upgraded and now covers the entire site collection. It can be customized as in the previous version but also includes Office Web Applications. The mobile view can be used with standard browsers as well. Documents can be viewed as text only or as image snapshots. In a demonstration I saw a comparison of a document at 432kb using up only 14kb when viewed in mobile view.
• Windows 7 Branch Cache
  This works as a content proxy for any asset such as documents and images. Branch caching can work in peer-to-peer mode, where computers in the workgroup are queried for a document before going to the server. You can also install a dedicated branch cache server using Windows 2008 R2 on the network. The branch cache is again diff aware and can greatly increase performance on satellite worker hubs.