-
Notifications
You must be signed in to change notification settings - Fork 6
Benchmarks
Flavio Serrapica edited this page Jul 12, 2016
·
40 revisions
In this section will be shown some benckmarks used for testing DMASON. For each of them will be provided a description and the reasons. Each test was running on a cluster with each node has the following features:
-
Processors: 2 x Intel(R) Xeon E5-2680, 8C, 2.7GHz, 20M Cache
-
Memory: 16GB DDR3
-
Hard Drive: 1TB
The below fields describe a simulation configuration:
- AOI: Area of interest
- NumAgents: Number of agents to simulate
- Width: Simulation field width
- Heigth: Simulation field heigth
- Rows: Rows number used for simulation field partitioning
- Columns: Columns number used for simulation field partitioning
- Mode: Indicates the kind of partitioning to use
This benchmark is useful for pushing the system to the extreme and measuring the its capability when it is used for a massive amount of work.
This banckmark has been performed wih the following configuration using an uniform partitioning:
- AOI: 10
- NumAgents: 3 Milions
- Width: 1000
- Heigth: 1000
Both rows and colums value changes following this mapping:
LPs | Rows | Columns | LPs per node |
---|---|---|---|
4 | 2 | 2 | 1 |
8 | 2 | 4 | 2 |
16 | 4 | 4 | 1 |
32 | 4 | 8 | 2 |
64 | 8 | 8 | 4 |
128 | 16 | 8 | 8 |
Each simulation was performed for 10 minutes.