A Novel Task Allocation for Maximizing Reliability Considering Fault-Tolerant in VANET ... 学习笔记

版权声明:本文为博主原创文章,转载请注明出处。 https://blog.csdn.net/gongxifacai_believe/article/details/82821218

A Novel Task Allocation for Maximizing Reliability Considering Fault-Tolerant in VANET Real Time Systems 学习笔记

1、这篇文章结合SDN和Fog Computing提出了一个HSFV体系结构。用指数分布建模,并将建模结果表示成一个最优化问题,然后用粒子群算法进行求解。
HSFV——hybrid SDN/Fog computing VANET architecture
FPSO-RO——fault-tolerant reliability optimization particle swarm optimization algorithm
2、SDN的作用:leverage centralized control, flexibility, programmability of SDN
Fog Computing的作用:ultra-low delay, location awareness of fog computing
体系结构:

本人认为:在这篇文章中,通篇考虑,SDN和雾计算不用也可以实现最优化任务分配,但体系结构可以参考。
3、In heterogeneous distributed network, there exits many solutions to optimize reliability. For example, traditional redundancy technology was extensively used in distributed network for fault tolerant[10], including hardware redundancy(e.g. a processing node can comprise multiple processors) and software redundancy(task replication among processing nodes).
在异构的分布式网络中,存在许多解决方案来优化可靠性。例如,传统的冗余技术在分布式网络中大量地用于差错容忍中,包括硬件冗余(例如,一个处理节点可以由多个处理器组成)和软件冗余(处理节点中的任务副本数)。
However, in practice, the real-time distributed system is vulnerable to reliability issues due to inevitable disturbances, such as hardware failure, software failure, and link failure( due to packet collision, poor channel conditions, vehicular high-mobility, etc. )
然而,实际上,实时分布式系统容易有可靠性不高的问题,因为它不可避免会损坏,比如硬件失败,软件失败和链路失败(由于数据包的碰撞,信道条件太差,车辆高速移动等等)。
4、假设:
(1)任务可以被任意分片;
(2)任务片之间没有依赖关系;
(3)任务片之间没有时间序列关系;
(4)每个分配的任务片可以在雾节点上立即执行;
(5)一个节点同时只可以处理一个任务片。
5、对任务的处理延迟和链路的通信延迟进行指数分布,指数分布的值为可靠性,求在处理延迟和通信延迟的和小于给定值的条件下,可靠性的最大化。这即是最优化问题。
The X = (x1, x2, …, xk) represents a task allocation, and xi ∈ X represents assigned task on fog node vi. The processing delay of xi is denotes as tc(xi, vi) = Q(xi)/Q(vi). The weight Q(xi) of subtask xi is its computational cost, and the weight Q(vi) of nodes vi is its processing capacity per unit time. Similarly, the communication delay of subtask xi is denotes as tτ (xi, evi,vj) = W(xi)/W(evi,vj). The weight W(xi) of subtask xi is its communication cost due to xi is needed to be transferred to node vi, and the weight W(e(vi,vj)) of link evi,vj is its communication capacity. i.e. data transfer rate.
X=(x1, x2, …, xk) 代表一次任务分配,xi ∈ X 代表在雾节点vi上分配的任务。xi的处理延迟表示为 tc(xi, vi) = Q(xi)/Q(vi)。Q(xi)是子任务xi的计算开销,Q(vi)是节点vi单位时间的处理能力。子任务xi的通信延迟表示为tτ (xi, evi,vj) = W(xi)/W(evi,vj)。子任务xi的W(xi)表示子任务xi传输到节点vi的通信开销,W(e(vi,vj))表示链路e(vi,vj)的通信能力,例如,数据传输率。
The entire application reliability is dependent on the reliability of system components(nodes, links). From work[13], under a task allocation X, the reliability of node vi is

during its processing delay tc(xi, vi), and the reliability of link e(vi,vj) is

during a communication delay tτ (xi, evi,vj). λi and εi represent the failure rate of processing node vi and communication link e(vi,vj) respectively. Then, the reliability Rxi of the subtask xi and the failure probability Fxi of xi can be formulated as:

and

respectively.
整个应用的可靠性依赖于系统组件(节点,链路)的可靠性。从文献[13]中得到,对于任务分配X,在处理延迟 tc(xi, vi)期间,节点vi的可靠性为

在通信延迟 tτ (xi, evi,vj)期间,链路e(vi,vj)的可靠性为

λi和εi分别代表处理节点vi和通信链路e(vi,vj)的失败率。子任务xi的可靠性Rxi和xi的失败概率可以分别形式化为



本人认为这篇文章公式的不足之处在于:
(1)对任务的处理延迟和链路的通信延迟是否满足指数分布还有待讨论;
(2)节点xi的可靠性是对于任务的处理延迟的指数分布值和链路的通信延迟的指数分布值进行简单相加,而我考虑,是否应该相加,相乘可不可以,如果相加,是否需要加上权重,而不是1:1的权重进行相加。
可借鉴之处:
可以列出一个子任务是否完成的可能性矩阵。
6、相关参数

系统参数 描述
Q(vi) 节点处理能力(MIPS) [140,300]
W 通信带宽(Mbps) [45,80]
λ 处理节点失败率 [0.001,0.003]
ε 通信链路失败率 [0.01,0.04]
OD 其他延迟 [0.003,0.010]

而文中这些值的来源依据并没有给出。
7、We also define the task complexity, i.e. divide the computational cost of task(MIPS) by the amount of task data(Mb).
定义任务的复杂度(task complexity),通过任务的数据量(Mb)来划分任务的计算开销。
8、As the future work, we plan to consider more failure factors e.g. software error and system constraint e.g. energy-consuming in our model.
未来工作,我们计划在模型中考虑更多的差错因素,比如软件上的失败,和系统约束,比如能量消耗。
9、这篇文章提到很重要的一篇参考文献:[13] S. M. Shatz, J. P. Wang, and M. Goto, “Task Allocation for Maximizing Reliability of Distributed Computer Systems,” IEEE Transactions on Computers, vol. 41, no. 9, pp. 1156-1168, Sep.1992.这是一篇发表在计算机分布式系统领域的CCF A类期刊上的一篇文章,是最初从分布式系统的角度考虑任务分配问题的文章,里面有关于任务分配的定义和泊松分布、指数分布的问题,下一步打算看一下这篇文章。

猜你喜欢

转载自blog.csdn.net/gongxifacai_believe/article/details/82821218