4. Typical Server Configurations
Microsoft often refers to the way that it
organizes servers working together as a topology, and the administrator
can modify the search application topology from within SharePoint's
Central Administration. Topology refers to the overall layout of the
servers and the software components on those servers.
The different ways in which servers can be
configured to provide scalability, performance, and redundancy for
SharePoint 2010 are vast and largely dependent on the particular
environment of each business. We will not attempt to give an exhaustive
example of each possible configuration but will attempt to outline the
basic models and the points where these may be safely and most
effectively modified to suit particular business needs.
As mentioned earlier, the most common and first
type of deployment any SharePoint administrator will encounter will be
the single server deployment, simply because this is the model where
all testing and development will start. However, few organizations with
serious document management and collaboration needs will utilize a
single server in a production environment. Even small, departmental
deployments will usually have at least a separate database server.
Single Server
A single server implementation for very small
corpora can be either part of an existing SharePoint farm or a
stand-alone search server with crawl and query components on it. It may
or may not have database components on the server. Most organizations
will have an existing database cluster that they will utilize for
performance and redundancy. Figure 3 shows a simple but typical single search server implementation with or without a database.
Figure 3. A single server deployment
Small Farm
Many organizations will demand more performance
and utilize a small search farm. This farm participates in overall
SharePoint tasks as well as dedicated search tasks. One of the web
servers may be outside of Network Load Balancing (NLB) to provide
performance for indexing or may participate in content delivery. Both
web servers also perform the query server role. The content serving
portion of the farm will certainly have additional web servers. There
is a single crawl server. See Figure 4.
Figure 4. A small farm installation
Medium Farm
In a medium-sized farm, there should be at
least two servers with a shared query server role that can host the web
server role. One of these may or may not be excluded from NLB for
indexing performance. There are two crawl servers to provide redundancy
for crawling. There is a single dedicated database cluster for storing
the crawl database, property database, and Administration database. See
Figure 5.
Figure 5. A medium search farm
Medium Dedicated Search Farm
At a certain point, the search components will
need to be separated from performing any content delivery tasks and
dedicated to search. In a dedicated search farm, there can be web
servers for delivering search requests and providing content to the
crawl server; however, they should be excluded from participating in
content delivery. Other components should be isolated to their own
servers and dedicated to their own roles. Combining roles is still
possible, but as the farm grows, dedicating the servers to individual
roles will be the wisest use of resources. See Figure 6.
Figure 6. A medium dedicated search farm
Large Dedicated Search Farm
For the largest deployments, all roles should
be hosted on multiple dedicated servers. Dedicated web servers for
indexing are still possible, but the focus should be on providing
high-performance query components to handle queries from a large
SharePoint farm. A large farm may have ten or more query servers, four
or more crawl servers, and at least two database clusters dedicated to
holding the search databases. As we saw in the SharePoint 2010
components section, index partitions can be separated across query
servers and mirrored evenly. See Figure 7.
Figure 7. A large dedicated search farm
