Hardware Sizing

How to Choose Hardware Resources

The hardware resources that are required to run Smart Previews depend on the following:

  • The size of your search environment
  • How frequently documents change in your environment
  • The Preview mode: Offline/Online

Note: Consult with your BA Insight representative for hardware sizing and configuration.

Assessing Your System Requirements

  • You should size your Smart Previews deployment to adequately handle ongoing operations and to build the Preview cache within the deployment time-frame that is acceptable to your organization.

  • The primary focus of your organization determines your requirements. See the topics below.

Deployment Time-Frame

  • You can reduce the time that is required for the initial cache build-up by scaling (adding more CPUs to the existing hardware).

  • You can also build-out (add servers to) your Smart Previews deployment.

  • The performance of Smart Previews increases with the number of cores that are available (whether you are scaling or building out).

  • In the case of scaling, you can reclaim the additional hardware once the cache is built.

Storage (Disk Space)

  • When deploying Smart Previews with less than optimum resources, sufficient temporary storage is required to accommodate crawled documents waiting to be processed.

  • For example, the typical plan is for approximately 60% of the largest content source.

Content Sources that Exceed 10M Documents

For large-scale installations (above ~10M documents), BA Insight recommends that you build a benchmarking exercise into the deployment project.

  • This exercise lets you measure the throughput in the target environment, on the target hardware, and with representative documents.
  • In this case, optimize the configuration to minimize the hardware footprint, while considering the characteristics of your environment and the desired performance.

Recommended Hardware Configurations

The following table describes some hardware guidelines for overall system performance:

Hardware Component

 Description
CPU

Smart Previews can support up to 32 cores, on a single VM.

  • If more CPUs are required, you can scale by adding additional servers.
  • The load on the hard drive increases as the number of CPUs or cores that are used by Smart Previews, increases.
  • On a 32-core system, Smart Previews attempts to process 32 files simultaneously.
  • For this reason, BA Insight recommends that you use high performance, high-speed hard drives with low access time.
  • This configuration enables maximum throughput.
  • For more information, see the Server Capacity row below.
Memory

Smart Previews expects 1 GB of memory per core for optimal results.

  • If you have a server with 32 cores, you optimally have 32 GB of memory available to Smart Previews.
  • Low latency and high frequency memory are preferred.
Local storage

BA Insight recommends that you use multiple disks for most environments.

  • Allocate 1 drive for the operating system.
  • Allocate a 2nd drive for the application component and logs.
  • To increase performance an additional drive can be used for the Smart Previews application folders incoming and temp.
Database servers

Databases on the SQL server have size requirements that are approximately 60% of the indexed content size.

  • For example, if you plan to index approximately 1 TB of data, allocate 600 GB of storage for your Smart Previews databases.
Note: Your BA Insight representative can assist you with this planning.

There are the 3 Smart Previews databases:

  • Preview Cache: The preview cache consumes more than 90% of the required space.
    • For larger environments you can scale your database capacity by creating new preview cache databases.
    • As a rule of thumb, plan to create a new preview cache database for every 5 million previews generated.
  • Preview Configuration
  • Preview User Profile
For information about how to configure multiple preview databases, see How to Manage Your Databases.

Server capacity

  • Configure Smart Previews to use all the available resources when the application is deployed on dedicated servers.

  • Maximum CPU utilization is approximately 90% during the preview generation process.

Note: If the server is not dedicated solely to preview generation, CPU utilization can be set to either:
  • Partially reduced (50%)
  • Reduced (30%)
Online preview

Preview response time is affected by the following:

  • file size
  • file type