Avoid network back pressure problem
Abstract : In this paper, we focus on the issue of stability in multihop wireless networks under flow-level dynamics, and explore the inefficiency and instability of the celebrated BackPressure algorithms. It has been well-known that the BackPressure (or MaxWeight) algorithms achieve queue stability and throughput optimality in a wide variety of scenarios. Yet, these results all rely on the assumptions that the set of flows is fixed, and that all the flows are long-lived and keep injecting packets into the network. Recently, in the presence of flow-level dynamics, where flows arrive and request to transmit a finite amount of packets, it has been shown that the MaxWeight algorithms may not guarantee stability due to channel fading or inefficient spatial reuse. However, these observations are made only for single-hop traffic, and thus have resulted in partial solutions that are limited to the single-hop scenarios. An interesting question is whether straightforward extensions of the previous solutions to the known instability problems would achieve throughput optimality in multihop traffic setting. To answer the question, we explore potential inefficiency and instability of the BackPressure algorithms, and provide interesting examples that are useful to obtain insights into developing an optimal solution. We also conduct simulations to further illustrate the instability issue of the Back-Pressure algorithms in various scenarios. Our study reveals that new types of inefficiencies may arise in the settings with multihop traffic due to underutilization of the link capacity or inefficient routing, and the stability problem becomes more challenging than in the single-hop traffic counterpart.
? Some times each and individual project estimates the additional relief loads they will put into the existing flare system and compare with the available capacity in the flare system.
? In most of the cases the new sources (PSVs & EBVs) are added to the existing flare system without any modification or upgrade of current system.
? This is the maximum pressure that can exist at the outlet of the device (source) without affecting its capacity.
? The process parameters shall be noted during blowdown period with the help of existing flow/pressure transmitters/gauges or temporarily installed instruments.
? It is a very interesting problem to study the stability issue of the Back-Pressure algorithms in the presence of flowlevel dynamics, and to identify potential causes of inefficiency in a more general setting.
? The Static Randomized scheduling algorithms that require knowledge of the arrival rates, a possible solution to the instability problem is to use per-destination queues at each node.
? Although flow-level dynamics plays a critical role in the instability problem, there are many different types of inefficiencies that cause instability in multihop traffic scenarios.
? The instability issue comes from underutilization of the link capacity or inefficient routing due to insufficient paths information.
• The back-pressure control proposed in this paper(BP) requires such loop detectors and an estimation of the total number of vehicles queuing at each node (gathering all possible directions).
• It is an extension of the algorithm proposed in Varaiya (2009) where internal/exit links are not differentiated, because exits may occur at any link of the network.
• Back-pressure control proposed in requires complete knowledge of the queues lengths matrix Q(t) and the routing rates.
• We propose to implement the two back-pressure controllers and to compare their behaviour.
? Their clever counterexamples show that in a wireless downlink system with time-varying link rates, the MaxWeight algorithm may fail to achieve the optimal throughput performance.
? We focus on a wireless downlink system where relay-assisted 2-hop communications can be adopted to improve throughput performance.
? We compare the performance of Q-BP and the Static Randomized scheduling algorithm.
? we also illustrate the instability issue with a numerical experiment that compares the scheduling performance of the Back-Pressure algorithms and that of a stable algorithm.
? This makes it difficult to develop a unified solution that achieves the optimal throughput performance in a general network setting.
|