On-Prem Component Metrics for Scaling and Management

Updated 7 hours ago by Michael Cretzman

During periods of high volume, you might want to autoscale On-Prem components up, and during periods of low volume, autoscale down.

To do this with confidence, you will want to know the performance range for On-Prem components.

This topic provides guidance on scaling On-Prem components such as the Learning Engine (LE) and Timescale DB.

Summary

Test

Configuration

Result

Workflow Setup - Timeseries (APM)

1 instance - 4CPU and 2GB Memory

20 parallel workflow executions can run without delay

Workflow Setup - Log Analysis

1 instance - 4CPU and 2GB Memory

6 parallel workflow executions can run without delay

24/7 Service Guard Setup - Timeseries (APM)

1 instance - 4CPU and 2GB Memory

30 24/7 Service Guard configs can be supported

24/7 Service Guard Setup - Log Analysis

1 instance - 4CPU and 2GB Memory

20 24/7 Service Guard configs can be supported

Mixed Setup and Workflows

1 instance - 4CPU and 2GB Memory

The following can run successfully with out any delay:

  • 3 APM Workflow
  • 4 Log analysis Workflow
  • 3 APM 24/7 Service Guard
  • 4 Log analysis 24/7 Service Guard

Test Results

Scenario 1

  • Configuration: 1 LE instance with 8CPU and 2GB Memory
  • Workflow setup: 8 time series 7 Log analysis running in parallel every 20 minutes
  • 24/7 Service Guard Setup: 8 time series 7 Log analysis

Result

  • Workflow Execution: No Delay
  • 24/7 Service Guard execution: Delay > 1 hr
  • No of LE analysis task in Queue: > 500
  • Over All status: Failed

Scenario 2

  • Configuration: 1 LE instance with 8CPU and 2GB Memory
  • Workflow setup : 4 time series 5 Log analysis running in parallel every 20 minutes
  • 24/7 Service Guard Setup: 4 time series 5 Log analysis

Result

  • Workflow Execution: No Delay
  • 24/7 Service Guard execution: Delay > 1 hr
  • No of LE analysis task in Queue: > 300
  • Over All status: Failed

Scenario 3

  • Configuration: 1 LE instance with 8CPU and 2GB Memory
  • Workflow setup: 3 time series 4 Log analysis running in parallel every 20 minutes
  • 24/7 Service Guard Setup: 3 time series 4 Log analysis

Result

  • Workflow Execution: No Delay
  • 24/7 Service Guard execution: No Delay
  • No of LE analysis task in Queue: < 20
  • Over All status: Passed

Scenario 4

  • Configuration: 1 LE instance with 4 CPU and 2GB Memory
  • Workflow setup: 3 time series 4 Log analysis running in parallel every 20 minutes
  • 24/7 Service Guard Setup: 3 time series 4 Log analysis

Result

  • Workflow Execution: No Delay
  • 24/7 Service Guard execution: No Delay
  • No of LE analysis task in Queue: < 20
  • Over All status: Passed

Scenario 5 - Horizontal Scaling

  • Configuration: 1 LE instance with 4 CPU and 2GB Memory
  • Workflow setup: 6 time series 8 Log analysis running in parallel every 20 minutes
  • 24/7 Service Guard Setup: 6 time series 8 Log analysis

Result

  • Workflow Execution: No Delay
  • 24/7 Service Guard execution: No Delay
  • No of LE analysis task in Queue: < 20
  • Over All status: Passed


How did we do?