Estimate performance and capacity requirements for SharePoint Foundation 2010 Search

Applies to: SharePoint Foundation 2010

This article discusses capacity planning information for a deployment of Microsoft SharePoint Foundation 2010 search.

In this article:

• Introduction

• Test farm characteristics

• Health and performance data

• Test results

• Recommendations

• Troubleshooting performance and scalability

Introduction

This article describes a specific deployment of SharePoint Foundation 2010 and includes the following specifications that were used during the performance and capacity testing of SharePoint Foundation 2010 :

• Test environment specifications, such as hardware, farm topology, and configuration

• The workload used for data generation that includes the number and class of users, and farm usage characteristics

• Test farm dataset, which includes database contents, search indexes, and external data sources

• Health and performance data that is specific to the tested environment

This article also contains common test data and recommendations for how to determine the hardware, topology, and configuration that you need to deploy a similar environment, and how to optimize the environment for appropriate capacity and performance characteristics.

SharePoint Foundation 2010 Search provides basic search functionality for local SharePoint Foundation content. If you need more advanced search functionality, such as indexing local shared folders and remote SharePoint Foundation or HTTP content, use Microsoft Search Server Express 2010 or Microsoft SharePoint Server 2010 search.

This article describes how to do the following:

• Define performance and capacity targets for the environment.

• Plan the hardware that is required to support the number of users and the features that you intend to deploy.

• Design the physical and logical topology for optimal reliability and efficiency.

• Test, validate, and scale up or scale out the environment to achieve performance and capacity targets. To increase the capacity and performance of one of the starting-point topologies, you can do one of two things. You can either scale up by increasing the capacity of the existing servers or scale out by adding additional servers to the topology.

• Monitor the environment for key indicators.

Test farm characteristics

This section describes the scenario of the test; the dataset that was used during the testing; the workloads that were put on SharePoint Foundation during performance gathering; the hardware that was used during the testing; and the topology for how that hardware was deployed.

This SharePoint Foundation Search scenario describes a typical search deployment, with the following characteristics:

• Search is provided for the SharePoint Foundation content in the farm. No external content can be added to the index (for example, intranet Web sites or file systems).

• Search can be run on only one server in the farm at a time. Scaling out is not supported.

• Limited system administration is available for search.

Dataset

This section describes the test farm dataset, which includes database contents and sizes, search indexes, and external data sources.

Hardware settings and topology

This section describes the kinds of computer hardware that we used in our lab and the farm configuration topology that we used in our tests.

Lab hardware and topology

This section describes the hardware that was used for testing.

Application server and content server

Microsoft SQL Server computer

Topology

The test farm is a two-server farm. One server is the Web and application server, and the second server is the database server.

Workload

This section describes the workload used for data generation. Workload characteristics include the number of users and farm usage characteristics.

Health and performance data

This section provides health and performance data that is specific to the test environment.

Query performance data

The following measurements were taken with 140,000 items in the search index. The columns give the measurements taken during the specific test. The results are at the bottom of the table. The CPU memory and IOPs were read directly from the Perfmon counters. The latency was measured by using a Microsoft Visual Studio Team System 2008 Database Edition project, which executed a set of approximately 2,000 queries on a team site.

Crawl performance data

The following measurements were taken during full crawls of the given content source. (Content source size is given in millions of items.) The columns show the measurements taken during the specific crawl, and the results are at the bottom of the table.

Test results

This section provides test data that show how the farm performed under load at different index sizes.

Query latency

The following graph displays the query latency percentiles for this farm as user load increases. A query percentile of 95 percent means that 95 percent of the query latencies measured were less than that value.

From this graph, you can see that increasing the size of the content crawled by this farm will affect query latency as the user load exceeds 20 concurrent users.

With approximately 5 million items in the index, the query latencies are very high, as shown in the following graph.

Query throughput

The following graph displays the query throughput for this farm as user load increases.

This graph shows that increasing the size of the content crawled by this farm will affect query throughput significantly, increasing it as the load exceeds about 10 concurrent users.

With approximately 5 million items in index, query throughput may be severely decreased as shown in the following graph.

Crawl rate

The following graph displays the crawl rate for this farm during the index acquisition stage at different existing index sizes. Notice that the crawl speed decreases as the index grows larger.

Recommendations

This section describes specific actions that you can take to optimize the environment for appropriate capacity and performance characteristics.

Hardware recommendations

For specific information about minimum and recommended system requirements, see Hardware and software requirements (SharePoint Foundation 2010).

Scaled-up and scaled-out topologies

The search subsystem for SharePoint Foundation 2010 cannot be scaled out. In general, we recommend that if there are more than 1 million items to be indexed, you should use Microsoft Search Server 2010 or SharePoint Server 2010 Search.

Some of the problems with low query throughput and high latency can be avoided by scaling up the CPU, RAM, and IOPS on the application servers and SQL Server computers.

Troubleshooting performance and scalability

To help you determine when you should scale up or scale out the system, use performance counters to monitor the health of the system.

Performance monitoring

Use the information in the following tables to determine which performance counters to monitor, and to which process the performance counters should be applied.

Web and application servers

The following table shows performance counters and processes to monitor for Web servers in the farm.

Database servers

The following table shows performance counters and processes to monitor for database servers in the farm.

About the author

Brion Stone is a Senior Program Manager for SharePoint Server search at Microsoft.