基于共享资源量的动态多资源公平分配策略

Dynamic fair allocation of multi-resources based on shared resource quantity

针对云计算共享系统中多资源分配问题,提出一种基于共享资源量的动态多资源公平分配策略。该策略根据不同用户资源需求和共享资源量建立一个线性规划模型,同时证明该模型满足公平分配的4个重要属性:动态帕累托最优、激励共享、动态无嫉妒性和防止策略性操作,而且给出一种改进的动态多资源公平分配算法来提高算法运行效率。实验结果表明,所提动态多资源公平分配策略能够在满足任务资源需求的同时,尽可能保证公平分配下最大化占优资源份额,并且改进的分配算法能够有效地提高资源的分配效率。

A dynamic fair allocation of multi-resources was proposed based on shared resource quantity for multi-resoures allocation problem in cloud shared computing system. Firstly, a linear programming model was given based on resource requirements and quantity of shared resource and this model was further proved which satisfies four fairness properties such as DPO, SI, DEF and SP. Secondly, an improved dynamic multi-resources fair allocation algorithm was introduced for the allocation efficiency. Finally, theoretical analysis and experiments demonstrate that this strategy can satisfy the demands as well as maximize the dominant share on the base of approaching fairness and the improved algorithm increases the allocation efficiency in the dynamic system.