Progressive or Conservative: Rationally Allocate Cooperative Work in Mobile Social Networks
Progressive or Conservative: Rationally Allocate Cooperative Work in Mobile Social Networks
复制标题
DOI：
10.1109/tpds.2014.2330298
复制
发表时间：
2015-07
影响因子：
5.3
通讯作者：
Wei Chang;Jie Wu
中科院分区：
文献类型：
--
作者：
Wei Chang;Jie Wu
互联网上有大量空闲的计算资源，这些资源可以用来完成巨大的任务。越来越多的应用程序被设计用于探索这些空闲资源。本文研究了移动的社交网络（MSNs）中的闲置计算资源，包括人类智能和机器计算能力。基于MSN的独特性，我们设计了一个新的合作系统，称为社会众包。该系统的分布式和无基础设施的特点使其比传统的众包平台更具吸引力。在所提出的系统中，一个巨大的工作是逐渐划分成较小的块，并从节点传播到节点。然而，如何划分和分配这些段是一个关键问题，它直接影响到工作的完成时间和系统吞吐量。由于缺乏全局信息，独立的中继节点可能会做出冲突的决定，这将导致参与节点上的工作负载分配不平衡。在本文中，我们发现，对于一个工作在不同的处理阶段，应该采取不同的工作量交换方案，从渐进的方法到保守的。在此基础上，我们提出了一种自适应的工作负载分配方案，在该方案中，参与节点可以根据相邻节点的工作负载状态逐渐切换其决策策略。通过使用我们的方法，系统的吞吐量可以显着提高，大型作品可以在接近最佳的时间内完成。与传统的调度问题不同，我们考虑了人类的拒绝，联系延迟和社会相似性。大量的仿真结果表明，我们提出的算法可以成功地充分利用在MSN中的空闲资源。
There are plenty of idle computational resources on the Internet, which could potentially be used for accomplishing huge tasks. More and more applications are being designed for exploring those idle resources. In this paper, we focus on the idle computational resources, including both human intelligence and machine computing abilities, in mobile social networks (MSNs). Based on the unique features of MSN, we design a new cooperative system, called social-crowdsourcing. The distributed and infrastructure-free features of the system make it more attractive than traditional crowdsourcing platforms. In the proposed system, a huge work is gradually partitioned into smaller pieces, and is propagated from node to node. However, how to partition and allocate these segments is a critical problem, which directly affects the work's completion time and system throughput. Due to the lack of global information, independent relay nodes are likely to make conflicted decisions, which will cause an unbalanced workload distribution on participating nodes. In this paper, we find that, for a work at different processing stages, one should adopt distinct workload exchanging schemes, moving from a progressive method to a conservative one. Based on this observation, we propose an adaptive workload allocation scheme, in which a participating node can gradually switch his decision strategy according to the workload statuses of neighboring nodes. By using our approach, system throughput can be significantly improved, and large works can finish within a nearly optimal time. Unlike in traditional scheduling problems, we take a human's rejection, contact delay, and social similarity into consideration. Extensive simulation results show that our proposed algorithms can successfully make full use of the idle resources in MSNs.