Performance Data for Lasernet Keep

This article provides performance data for Lasernet Keep 11.0 and shows its processing speed during specific tasks.

Note

Actual performance of your Keep system may differ from the performance tests due to factors such as hardware specification and software configuration. Performance data is provided for illustrative purposes only.

Test Environment

Performance tests used the following:

• Hardware

• Test configuration

• System size

Hardware

Testing was performed on dedicated hardware and used two hosts with the following configuration:

• 1 x Intel i7 8700K 3.7GHz (4.7GHz Turbo) 6 Cores (12 Threads)

• 64 GB RAM

• SSD drives

• 1 GB network adapter

Test Configuration

The two hosts ran virtual machines under the following configuration:

Host 1

Host 2

System Size

Tests measured the performance of the following Keep system sizes. Although the number of documents is the main factor that influences performance, the number of document definitions and key definitions are also major factors.

• Small: 5 document definitions, 20 key definitions, 3 line items, 10 million documents

• Medium: 10 document definitions, 30 key definitions, 3 line items, 50 million documents

• Large: 20 document definitions, 50 key definitions, 3 line items, 200 million documents

Input Data Size

The performance tests used the following input data sizes:

• 0 KB: Metadata only (no attachment)

• 100 KB: Metadata and PDF document attachment

• 200 KB: Metadata and PDF document attachment

• 1 MB: Metadata and PDF document attachment

• 2 MB: Metadata and PDF document attachment

Test Standards and Measurements

The tests used Keep REST API endpoints. Each combination shown in the following data tables is considered to be a distinct test. For example:

• Small system standalone — 100 KB input data.

• Medium system standalone — 1 MB input data.

We ran each test three times under the stated hardware configuration, using the same number of threads, and then calculated the average of the test’s results. The tests used 10 simultaneous threads to simulate realistic usage.

We endeavored to eliminate temporary environmental problems. For example, we paused Windows updates during testing to prevent background processes from affecting CPU load.

We measured throughput and processing time:

• Throughput was measured in actions per second (a/s).

• Processing time was measured in seconds.

Test Results

The following sections contain system performance results for the specified scenarios, using 10 simultaneous threads to create, get, or search documents:

“Create” and “get” tests used five input data sizes:

• 0 KB

• 100 KB

• 200 KB

• 1 MB

• 2 MB

“Search” tests used three data sizes (result per page/total results):

• 1/1

• 0/100

• 25/1000

Create Documents Throughput (Actions Per Second)

Get Documents Throughput (Actions Per Second)

Search Documents Throughput (Actions Per Second)