Your friends are more powerful than you: Efficient task offloading through social contacts
Your friends are more powerful than you: Efficient task offloading through social contacts
复制标题
DOI：
10.1109/icc.2014.6883300
复制
发表时间：
2014-06
期刊：
影响因子：
--
通讯作者：
Qingyu Li;Panlong Yang;Yubo Yan;Yue Tao
中科院分区：
文献类型：
--
作者：
Qingyu Li;Panlong Yang;Yubo Yan;Yue Tao
在这项工作中，我们研究了移动社交网络中的分布式和平衡的任务重新分配。先前的研究已经通过随机行走模型展示了随机选择在负载平衡中的威力。受到“球和仓”理论中“2 选择”范式的启发，我们使用真实跟踪数据“MobiClique”评估了这个简单但有效的方案。根据初步评估结果，我们发现，社会关系与纯随机游走模型显着不同，会给任务重新分配带来以下挑战：首先，友谊相对稳定，这会导致任务分配不平衡。其次，有些用户见面的频率很低，这会导致难以忍受的时间延迟和任务分配不均匀。为了应对这些挑战，我们提出了“iTop-K”，利用基本概念，即你的朋友比你更强大，鼓励移动用户在亲密的朋友之间分配任务，而不是纯粹的随机分配。通过选择“top-K”好友，我们可以同时实现负载均衡和保证网络性能。实验研究验证了我们的方案并显示了有效性。在典型的工作场景中，应用真实轨迹驱动的模拟，我们的性能比传统的随机选择高出 15 倍，而没有优先级方法的社会关系分配则高出 9 倍。
In this work, we investigate the distributed and balanced task reassignment in mobile social networks. Previous studies have shown the power of random choice in load balancing with random walking model. Inspired by the `2-choice' paradigm in `ball and bin' theory, we evaluate this simple but effective scheme with real trace data `MobiClique'. According to the preliminary evaluation results, we find that, social relationship significantly differs from pure random walk model, and will bring challenges in task reassignment in the followings: First, friendships are relatively stable, which will lead to imbalanced task assignment. Second, some users meet quite infrequently, which will lead to intolerable time delay and uneven task distribution. In tackling with these challenges, we propose `iTop-K', leveraging the basic concept, i.e., your friends are more powerful than you, which encourages mobile users to assign tasks among intimate friends instead of pure random assignment. With the selection of `top-K' friends, we can achieve load balancing and guaranteed network performance at the same time. Experimental studies verify our scheme and show the effectiveness. In typical working scenario, where real-trace driven simulation is applied, ours outperforms the conventional random choice up to 15×, and the social relationship assignment without priority method up to 9×.