SharePoint 2010 Components : Web Server Role, Query Server Role, Database Server Role

SharePoint 2010 has a number of performance and redundancy features. The search capabilities have been redesigned to allow for a broader ability to scale and more points for redundancy.

The new architecture for SharePoint 2010 provides a more compartmentalized approach to search by dividing the tasks that the search mechanism performs into different roles that can also be spread out across physical or virtual servers, as well as further divisions within these roles. The four server roles for search are as follows:

• Web server role
• Query server role
• Crawl server role
• Database server role

The query server and crawl server roles are unique to the search component, whereas the web server and database server roles can be utilized by and are necessary for other components of SharePoint 2010.

1. Web Server Role

Servers hosting the web server role host the web components of SharePoint 2010 that provide the user interface for searching. These components, such as search center sites, Web Parts, and web pages that host query boxes and result pages, are delivered from servers with the web server role to the end users. These components send requests to servers hosting the query server role and receive and display the result set.

The web server role may not be necessary in SharePoint farms that are dedicated for search, as other farms that are utilizing the search farm will handle this role and communicate with the search farm directly from their web servers. The web server role is often combined in smaller deployments with web servers serving content or with other search server roles.

2. Query Server Role

The query server role serves results to web servers. Query servers receive requests from servers with the web server role and forward these requests to all servers in a farm with the query server role. They then process the query against all index partitions and return their results to the requesting server, which then forwards the results to the requesting web server.

On each query server, there is a query processor, which trims the result set for security, detects duplicates, and assigns the appropriate associated properties to each result from the property store. Any SharePoint farm providing search must have at least one server hosting the query server role. However, a farm may call search from another farm and therefore not need the query server role.

The query server role, like other application roles in SharePoint, can be hosted on a server with other application server roles. This makes SharePoint 2010 very versatile but may cause confusion when planning resource usage. Having all servers provide all roles is not optimal resource usage, as some demanding roles may cause other roles to perform poorly. Caution and consideration regarding the role and demand of each server and each task are therefore advised.

The query server holds the index on its file structure or a file structure relative to it. A query server can host either the entire index or index partitions—sections of the index that can be assigned to different query servers by the administrator for load, performance, and redundancy. Index partitions may be duplicated on a number of servers with the query server role to provide redundancy. Adding query servers with the index partitioned across those query servers will also increase search query performance and reduce result latency.

Index Partitions

An index partition is a portion of the entire search index. Microsoft has designed the index to be broken into logical sections that can be distributed and mirrored across query servers. Generally, index partitions are spread across servers and represent an equal amount of crawled data. Indexes may also be partitioned on a single server. They can also be mirrored on another server or set of servers to provide redundancy.

Imagine, for example, that a SharePoint farm has 300GB of crawled data and three query servers. Each query server can hold a single index partition representing 100GB of crawled data. Query speed is increased because the load of searching the index is distributed over servers and divided by three. The query servers take time to look into the index for any given query, and therefore searching in smaller partitions across multiple servers is substantially more performant. An additional mirror of each partition can also be added to each query server to insure redundancy. Should any one query server fail, the remaining query servers still have all portions of the index and can continue to serve results. See Figure 1.

Figure 1. Three query servers with mirrored partitions

Crawl Server Role

The crawl server role is responsible for crawling content. This crawling mechanism is similar to other web crawling technologies, except that it is specifically designed to crawl and index SharePoint content, including user profiles from a directory, associated document metadata, custom properties, file shares, Exchange public folders, web content, and database and custom content through the BSC (as well as content via iFilters and protocol handlers).

The crawl servers host the crawler components, and, like the query server role, at least one server in a SharePoint 2010 farm providing search must host the crawl server role. Crawlers on the crawl servers are associated with crawl databases. Each crawler is associated with one crawl database.

It is recommended that the Search Administration component also be hosted on the server with the crawl server role. However, it can be hosted on any server in the farm. SharePoint 2010 hosts only a single Search Administration component per Search service application.

Search Service Application (SSA)

SharePoint 2010 has its core services broken into service applications. These applications, which deliver much of the functionality of SharePoint 2010, are separated to provide granularity and scalability when managing many of the different features available in SharePoint 2010. These services include but are not limited to the User Profile service, the Business Data Connectivity service, the Managed Metadata service, and the Search service, among others. Additionally, third-party vendors or solution providers could provide custom service applications that plug into SharePoint 2010, although at the time of writing, there were not any good examples of a third-party service application.

The Search service application is the service application that is responsible for the search engine. It manages the crawler and the indexes as well as any modifications to topology or search functionality at the index level.

3. Database Server Role

In a SharePoint 2010 Search deployment, the search databases are hosted on a server with the database server role. It is also possible to host other SharePoint 2010 databases on the same server or separate search and content database roles. Servers with the database server role can be mirrored or clustered to provide redundancy.

There are three types of databases utilized by a SharePoint 2010 farm providing search: property databases, crawl databases, and Search Administration databases.

Aside from disk size and performance limitations, there are no other considerations that limit hosting other databases, such as SharePoint content databases, on a SharePoint 2010 server with the database server role.

• Property databases: Property databases hold property metadata for crawled items. These properties can be crawled document metadata or associated custom properties from SharePoint 2010.
• Crawl databases: Crawl databases store a history of the crawl. They also manage the crawl operations by indicating start and stop points. A single crawl database can have one or more crawlers associated with it. However, a single crawler can be associated with only one crawl database.
• Search Administration databases: Search Administration databases store search configuration data such as scopes and refiners and security information for the crawled content. Only one Search Administration database is permitted per Search service application.