DPS-CR: A Task Scheduling Algorithm Based on
Computation Reuse in Vehicular Edge Computing
Authors: Zhipeng Li,Shihong Hu,Bin Tang
Institute: Hohai University
Speaker: Zhipeng Li
Directory
▶01 Background and Significance
▶04 Solution
▶02 Data set
▶05 Performance Evaluation
▶03 Problem
▶06 Conclusion
1 Background and Significance
• The rapid development of mobile communication
and computing technology
•
Limitation of the computational capability of
mobile devices
• Cloud computing and vehicular edge
computing(VEC)
2 Data set
We simulate two different traffic models:
1.congested scenario
2.uncongested scenario
Uncongested scenario
Congested scenario
Intervals
t=1~2500
t=1~2500
Traffic model parameters
λ=1
λ=2
3 Problem
1.How to improve computational reusability by mining data similarity.
2.The tasks generated by vehicles typically have bandwidth limitations.
4 Solution
For tasks with spatiotemporal attributes, the inputs of
tasks with similar attributes tend to produce similar
outputs.
Task reusability assessment:
1. same vehicle position coordinates
2. same task attributes (bandwidth, workload, task
size)
Our goal is to make full use of bandwidth resources to reduce system cost as much as possible.
Thus, we can conclude that this is a 0-1 knapsack problem and is NP-hard.
Allocable bandwidth resource → backpack capacity
Bandwidth that needs to be allocated for task → volume of the item
System cost generated by the completion of the task → value
𝑑𝑝[𝑝][𝑞] = 𝑚𝑖𝑛(𝑑𝑝[𝑝 − 1][𝑞] + 𝑣𝑎𝑙𝑢𝑒[𝑖0],
𝑑𝑝[𝑝 − 1][𝑞 − 𝑛, 𝑖] + 𝑣𝑎𝑙𝑢𝑒[𝑖𝑚])
Combining Task Reusability Assessment and Dynamic Programming to form our DPS-CR algorithm.
5 Performance Evaluation
6 Conclusions
•Multi-Base Station Environment
•Bandwidth-Constrained Tasks
•Optimizing BS Storage Capacity and Spatiotemporal Task Reusability
THANKS FOR YOUR ATTENTIONS