Advances, Systems and Applications
Table 2 Reward values for different states
From: CLQLMRS: improving cache locality in MapReduce job scheduling using Q-learning
States | Rewards |
|---|---|
Cache-locality | (+ 1 × number of tasks with cache locality) / (total number of tasks) |
Data-locality | (+ 0.5 × number of tasks with data locality) / (total number of tasks) |
Cache rack-locality | (-0.25 × number of tasks with cache rack locality) / (total number of tasks) |
Data rack-locality | (-0.5 × number of tasks with data rack locality) / (total number of tasks) |
Nothing | (-1 × number of tasks with no locality) / (total number of tasks) |