Yu Hongfang

Traffic-aware VDC embedding in data center: A case study of fattree

Virtualization is a common technology for resource sharing in data center. To make efficient use of data center resources, the key challenge is to map customer demands (modeled as virtual data center, VDC) to the physical data center effectively. In this paper, we focus on this problem. Distinct with previous works, our study of VDC embedding problem is under the assumption that switch resource is the bottleneck of data center networks (DCNs). To this end, we not only propose relative cost to evaluate embedding strategy, decouple embedding problem into VM placement with marginal resource assignment and virtual link mapping with decided source-destination based on the property of fat-tree, but also design the traffic aware embedding algorithm (TAE) and first fit virtual link mapping (FFLM) to map virtual data center requests to a physical data center. Simulation results show that TAE+FFLM could increase acceptance rate and reduce network cost (about 49% in the case) at the same time. The traffic aware embedding algorithm reduces the load of core-link traffic and brings the optimization opportunity for data center network energy conservation.

Considering the utilization of idle time slots for DAG scheduling in optical grid applications

Cost-effective optical network solves the bottleneck of grid issues, thus promoting the study of optical grid. Task scheduling is the key optical grid research. List scheduling is one of the most common heuristic algorithms and the temporal and spatial mapping of a program to a parallel system. In this paper, we use the directed acyclic graph (DAG) to represent the tasks and the constraints among them. First we make some changes on the edge scheduling. Then by exploiting of unused time slots between edges or nodes which have been scheduled, the extended scheduling algorithm is proposed, which allows scheduling edge or node on the available time slots. Thus the edges and nodes can be scheduled within the first time. The extended algorithm takes the communication delay under consideration and can be used in heterogeneous environments. By way of evaluating the performance of the extended algorithm, it is conducive to DAG with higher degree of parallelism and is proved to be feasible for data-intensive applications.

Link weight design for green line-card model based IP networks

Reducing unnecessary energy consumption for networks attracts attention worldwide because of its potential economic benefits and environmental impact. Optimizing link weight is a useful and practical way to yield idler network devices that can produce opportunities for power saving. In this paper, we focus on designing link weight for green IP network with minimal energy consumption. We formulate the problem as a Mixed Integer Linear Programming (MILP) which is to minimize the whole power consumption by closing links and line-cards as many as possible based on link weight while satisfying certain network performances. Because the problem is NP-hard, we propose an efficient heuristics, called Static Weight Design Algorithm based on line-card (SWDA (line-card)), which sequentially determines the on/off state of links and line-cards based on rules to solve the problem in approximation way with low complexity. In addition, the link on/off state is essentially determined by link weight assignment. Simulation results reveal that the power savings of SWDA (line-card) can be significant and be close to that of MILP. Moreover, it will be more effective for networks with high node degrees and low traffic load.

Novel algorithm of finding good candidate pre-configuration cycles in survivable WDM mesh network

We present a novel algorithm of finding cycles, called the Fast Cycles Mining Algorithm (FCMA), for efficient p-cycle network design in WDM networks. The algorithm is also flexible in that the number and the length of cycles generated are controlled by several input parameters. The problem of wavelength assignment on p-cycle is considered in the algorithm. This algorithm is scalable and especially suitable for survivable WDM mesh networks. Finally, the performance of the algorithm is gauged by running on some real world network topologies.