Swarming elements benchmarks
Benchmarks
| # | Specification | Scope | Metric | Test Setup |
|---|---|---|---|---|
| 1 | Swarming 1 element | DMS | 0.591 s | DaaS cluster. |
| 2 | Swarming 10 elements | DMS | 1.590 s | DaaS cluster. |
| 3 | Swarming 100 elements | DMS | 9.670 s | DaaS cluster. |
| 4 | Raw swarming time for 1 element | DMS | 0.154 s | DaaS cluster. Measured swarm action only, without time needed to start the element. |
| 5 | Failover switch 100 elements | DMS | 50 s | Non-swarming system. Measured until Cube availability. Nothing else on the Failover Agent. |
| 6 | Swarming 100 elements (Failover-like switchover) | DMS | 10 s | On-premises STaaS. Measured until all swarm actions were completed. Elements became available one by one. |
| 7 | Element startup (100 elements present) (non-swarming) | DMS | 4.862 s | DaaS Cluster. Non-swarming system (loading element config from XML) |
| 8 | Element startup (100 elements present) | DMS | 5.256 s | DaaS Cluster. Swarming system (loading element config from database) |
| 9 | Setting element property | DMS | 0.024 s | Non-swarming, 10.5.2 |
| 10 | Setting element property | DMS | 0.025 s | Swarming-enabled dedicated clustered storage setup, 10.5.2 |
| 11 | Setting element property | DMS | 0.070 s | Swarming-enabled STaaS setup, 10.5.2 |
General Notes
- Benchmarks were taken with simple elements that do not require much loading of element data/alarm events from the database on element startup. The measurements mainly show the time needed to swarm an element aside from the element startup.
- Measured per element, swarming multiple elements is faster, as multiple elements can be swarmed simultaneously.
- Measurements are for swarm actions towards one Agent. Parallel swarm actions to multiple targets will provide better timing.