Overview

• The main idea of the test to understand the effect of horizontal scaling on Eureka by perform a CICO test of 20 users and 40 users on a non-HA and HA environment. Additional task is to check how availability zone affect the performance switching between us-east-1a and us-east-1b.PERF-1225: Effect of horizontal scaling on Eureka

Summary

• All tests completed successfully.

• Testing results in non-HA and HA environments show that non-HA mode perform better than HA

  • 20 vUsers tests response time slightly faster in non-HA 

    • CI 609ms - 17%(105ms), CO 1346ms - 9%(125ms)

  • 40 vUsers tests response time faster in non-HA 

    • CI 870ms - 54%(471ms), CO 2039ms - 23%(486ms)

• Testing results in different AZ (availability zone) reveal light performance improvement (3%) in 20 vUser tests in us-east-1a zone (the same database AZ) but for 40vUsers test improvement in us-east-1a was only in HA mode (5%). In non-HA response times better in us-east-1b 10% in average for both CI and CO.

• Additional test runs in ECS environment show that performance of CI/CO flows was not affected by horizontal scaling (non-HA, HA, 4 tasks). It perform the same or slightly better - 30-60 ms.

• Additional test results on ECS environment in HA and non-HA mode with 40 vUsers. This effort also include 2 tests on HA mode (2 tasks for CICO involved modules) and 2 tests with 4 tasks for CICO involved modules:

  • In a new environment comparing HA and non-HA we see that Check-In/Check-Out perform 5% better (CI - 28ms, CO - 58ms) in HA mode

• Raising number of tasks to 4 do not affect performance

Test Runs 

Results

Test results