Friday, July 02, 2010

Scalable SharePoint 2010 Farm & Services Architecture

This post describes a proposed solution design for providing an intranet collaboration farm capable to scale out to support a publishing portal and extranet collaboration with partners and suppliers. The design is based on a set of diverese sites types identified in an Information Architecture analysis, and the general planning recommendations for these types. In addition, several non-functional requirements is accounted for in the design, such as farm security, robustness and availability.

The SharePoint 2010 farm is designed with scale out support for different future solution areas. The initial farm architecture will be designed to support collaboration for internal users only, with an option to also add an intranet publishing solution. The long term goal of the farm is to eventually support collaboration with external users also, such as partners an suppliers.

All SharePoint 2010 solutions are built from a set of elements ranging from the farm hosting the solutions down to sites that contains the actual functionality and information of the solutions.

SharePoint solutions are deployed into SharePoint web-applications in a SharePoint farm. Web-applications are management containers for site collections. SharePoint uses site collections to structure the sites and subsites that implements the actual functionality of SharePoint solutions.

Site collections bridge logical architecture and information architecture (IA). The design goals for site collections in the model are to create logical divisions of content and functionality, in addition to satisfy requirements for URL design. To satisfy the requirements for URL design, each web application includes a single root-level site collection, such as ://portal/. In addition, a set of managed paths are used to incorporate a second tier of site collections, such as ://portal/HR/ and ://portal/sites/***.

All web-applications in a farm share a group of common SharePoint 2010 service applications that provide shared services such as indexing & search to the farm. I use the term “shared services provider” (SSP) in this post for such a group of service applications. A farm can contain multiple groups of shared services, such as an intranet SSP and an extranet SSP.

The overall SharePoint farm architecture for internal collaboration can be summarized like this (click to enlarge):

The initial SharePoint 2010 farm will be configured to support two solutions: providing team-sites for internal collaboration in groups or projects, and for providing My Sites for employees. These two web-applications will share a group of shared service applications (“default group”) for search, user profiles and enterprise taxonomy.

The User Profile service is required to provide My Sites. In addition, this service application is what provides all social features such as tagging, rating and social bookmarking.

Note that a search service application is needed for more than just crawling and queries, the social tagging functionality in SP2010 also requires search. The reason is that several web-parts such as Note Board, Tag Cloud and Tagged Items in some modes depends on the search service to do security trimming to provide their content. This includes the tag profile page, which use such web-parts. You can also use the search service and the content search, people search and social search capabilities when customizing the user experience of your solutions. A typical example is creating a tag cloud web-part for managed metadata and enterprise keywords columns for lists and libraries, not just for social bookmarking (it is not really social tagging that is implemented in SP2010) with the standard "Tags & Notes" ribbon.

If you use FAST Search for SharePoint 2010 (FS4SP) then the FAST Content SSA will provide the content sources that are crawled for search results, except for people search. The FAST Query SSA has a different set of content sources, including the one feeding the people scope. Only one Search Service Application (SSA) should be associated with your web-application, either FAST or standard SharePoint 2010 Server search. Note that FAST doesn't index the social tags, which affects those social tagging web-parts that depends on search (see above).

The intranet will use Active Directory as the identity provider for authentication in classic mode. The web-application authentication mode can later on be switched from classic to claims-based as needed. Note that it is not recommended to switch the mode from claims-based to classic.

The farm is also designed to scale-out to support future SharePoint 2010 solutions such as a publishing intranet (://puzzlepart). Note that other service applications not shown here might be required to support the functional solution design and future solutions.

The different web-applications in the farm runs in separate application pools to achieve process isolation. This way an error or crash in one application will be isolated and not affect the other web-applications in the farm.

Anywhere access for mobile employees to the intranet sites is provided using Forefront Unified Access Gateway (UAG). This supports a wide range of locations such as at home or on the road, and a wide range of devices from laptops to smart phones.

The overall SharePoint farm architecture target for collaboration with external partners and suppliers can be summarized like this (click to enlarge):

The extended 2011+ version of the farm is scaled-out to support two new solutions; one is the ://puzzlepart web-application for intranet publishing, the other is an extranet web-application for collaboration with external users such as partners and suppliers. The former connects to the intranet shared services provider (“default group”) and will as such reuse and share search, user profiles, and the enterprise taxonomy (metadata and content types) with the existing solutions for team-sites and my sites.

The extranet collaboration solution must not be connected to the intranet SSP for information security reasons. Using a separate group of shared services for all service applications that may contain or expose confidential information is strongly recommended – an extranet SSP (“custom group”). A typical service application of this kind is search. Using separate service applications prevents accidental information exposure due to e.g. misconfiguration of a service application.

The User Profile service application is included in the extranet SSP even if the extranet will not provide My Sites at all. This service is what enables user to do social tagging and bookmarking, and must thus be part of the extranet SSP to provide social features. It is not recommended to connect the extranet to the intranet User Profile service for the same reasons as for the Search service.

Still, some service applications are typically shared across both the intranet and the extranet web-applications. This includes the enterprise taxonomy – you need to provide for consistent classification and tagging of content across all solutions to drive findability. The managed metadata service, including content type hubs, is specifically built for the purpose of being both shared and syndicated across multiple solutions.

Another security aspect is access control; the extranet must use a separate claims-based authentication provider, that can be federated with external identity providers. The intranet and extranet web-applications must trust different providers for authorization. Using separate providers prevents accidental information exposure due to e.g. misconfiguration of site access control and group memberships. Note that multi-mode authentication is preferred over mixed mode for extranet collaboration.

A final aspect of the extended farm design is that the extranet web-application and shared services are run in separate application pools from the intranet. This ensures process isolation to prevent errors and crashes in the extranet processes to affect the intranet solutions. It also isolates malicious use of the extranet process resources from affecting the intranet processes, such as denial of service attacks.

Remote access for external users to the extranet collaboration web-application is provided using Forefront Unified Access Gateway (UAG) combined with ADFS for federated identity and access management (IAM). Using UAG saves you from setting up and managing a separate extranet farm in a DMZ perimeter zone; instead you give a controlled set of external users secure anywhere access to a controlled set of internal applications hosted on the intranet farm.

No comments: