Xiaoshan Yu

Distributed Flow Scheduling in Energy-Aware Data Center Networks

Recently, the strategy of powering off unneeded network devices is proposed for energy-aware data center networks (DCNs). However, efficient flow scheduling is required for balancing the load of numerous traffic flows in the irregular topologies. In this paper, we propose distributed flow scheduling (DFS) for energy-aware DCNs. DFS leverages multiple distributed schedulers and schedules elephant flows to suitable routes. Compared with previous solutions, DFS achieves ideal network performance with better scalability and efficiency. Finally, the evaluations demonstrate that DFS is a feasible system to be implemented in energy-aware DCNs.

A hybrid packet-circuit switched router for optical network on chip

The increasing number of Intellectual Property (IP) cores challenges the traditional electrical Network on Chip (NoC). Silicon nanophotonics becomes a leading technology because of offering several benefits for NoC. Also, On-chip services, including guaranteed service and best-effort service, have different traffic characteristics. This has an important influence on the performance of NoC. This paper proposes a hybrid packet-circuit switched router for optical Network on Chip (ONoC). It can support optical circuit switching (OCS) and optical packet switching (OPS) in parallel and simultaneously, in order to optimize the performance of the network with both services. According to the simulation results, the proposed architecture achieves lower latency and higher throughput than the traditional architectures in the same network scale.

Enhancing Performance of Cloud Computing Data Center Networks by Hybrid Switching Architecture

Cloud computing services have driven a new design of data center networks. Hybrid switching architecture is one of the promising solutions since it makes better tradeoff between the network performance and technical feasibility. However, as the existing hybrid networks only deploy one-hop optical circuit switching (OCS) in the top layer, the flexibility and scalability is limited. To address this problem, a distributed OCS model is proposed. To reduce the high blocking ratio, the WDM and SDM technologies are introduced to increase the connectivity of the optical network. Moreover a multi-wavelength optical switch based on microring resonators is designed to enable the fast switching. Based on this model, the multi-rooted tree based hybrid architecture with deep integration of optical connection is constructed. A new way to solve the mixed traffic scheduling problem is also provided by delivering the small flows and large flows through the different networks. The simulation results indicate that the multi-rooted tree based hybrid architecture achieves better performance under various traffic patterns. It also introduces less control overhead compared with the existing traffic scheduling schemes.

Energy-Aware on-chip virtual machine placement for cloud-supported cyber-physical systems

Abstract Recent trends in the design of cyber-physical systems (CPS) are moving towards heterogeneous multi-core architectures with cloud support. In this paper, we propose an energy-aware scheme for virtual machine placement in cloud-supported CPS with Network-on-Chip (NoC) architecture. We formulate the energy-aware on-chip virtual machine placement problem as an optimization problem, and design a heuristic scheme based on ant-colony optimization. We address problems of slow convergence speed and easily falling into stagnation in ant-colony algorithm by employing pheromone diffusion model that makes the proposed scheme more efficient. Simulation results show that our scheme achieves much higher energy efficiency compared with previous schemes with different network sizes and traffic models.

A scalable AWG-based data center network for cloud computing

With the development of cloud computing and other online applications, the traffic for data center network (DCN) has increased significantly. Therefore, it is extremely important for DCNs to support more and more servers and provide high scalability, high throughput and low latency. Some current topologies for data centers have such inherent problems as poor scalability, lack of path diversity, cabling complexity, etc. This paper proposes a scalable AWG-based optical interconnection network for data centers, which is called OIT. OIT possesses good scalability and path diversity and benefits from the inherent parallelism and high capacity of WDM and AWG, which makes it a suitable candidate topology for data centers in the cloud computing era. A multi-path routing algorithm is also designed to utilize OIT?s parallel links and distribute the load more evenly. The simulation results show that the packet latency and network throughput performance of OIT is better than that of fat tree topology under uniform random distribution or 50%, 80% intra pod traffic distribution and different packet sizes.

A crosstalk-aware wavelength assignment method for optical network-on-chip

Optical Network-on-Chip (ONoC) is a promising on-chip communication architecture for future many-core systems due to its high-performance and energy-efficient characteristics. However, the crosstalk problem is an obstacle to the realization of ONoC. In large scale ONoC, the crosstalk noise considerably reduces data transmission reliability via the signal-to-noise ratio (SNR), and limits the scalability of ONoC. In this letter, we propose a crosstalk-aware wavelength assignment (CWA) method to mitigate the crosstalk problem. Analyses results indicate that our CWA method significantly improves the worst-case SNR and allows a higher network scalability.

RingCube - An incrementally scale-out optical interconnect for cloud computing data center

With high capacity and low power consumption, the optical interconnect provides a promising solution to address the communication bottlenecks in cloud computing data centers. However, as the scale of data center keeps an explosive growth trend, the centralized optical switching architecture may suffer from scalability issues, such as the costly switching fabric, the complicated control system, and the inflexible expansion mode. A distributed optical interconnect named RingCube ( k , n , m ) is proposed to solve this problem. By embedding hypercube into ring topology and utilizing the multi-wavelength communication strategy, the architecture realizes incremental expansion at a fine granularity, without device replacement, port reservation or network rewiring. The proposed wavelength allocation strategies effectively reduce the number of wavelengths that is necessary to overcome the high blocking ratio of Optical Circuit Switching (OCS). Moreover a modular multi-wavelength optical switch structure is designed to realize the on-demand expansion of the wavelength switching capacity. The simulation indicates that RingCube achieves considerable performance and good scalability while maintains very low cost. We propose a scalable all-optical interconnect for cloud computing data center.A distributed wavelength-based Optical Circuit Switching is employed to build a path over multiple hops.The wavelength allocation strategy effectively enhances the connectivity and scalability of the network.Several approaches are taken to ensure a fast optical path setup.The optical switch enables an on demand extension of the wavelength switching capacity.

A new optical interconnection network for data centers

The network topology is the foundation of data center networks, and it is directly related to the performance, management, operation, maintenance and many other aspects of data center networks. With the rapid growth of cloud computing, the demand for high bandwidth and low network latency increases greatly. Traditional tree-based topologies of data centers have some inherent problems such as poor scalability, the lack of path diversity, low bisection bandwidth, etc. In this paper, we propose a new optical interconnection topology for data centers, which is called OIT. In our theoretical analysis, OIT has good characters of scalability and path diversity. We also design a new routing algorithm according to its network features.

Cell-Tree: A cost-efficient and flat network topology for data centers

The data center is an important infrastructure for cloud service. With the rapid development of cloud computing, numerous challenges such as cross-platform requirements, network bandwidth, virtual machine migration, privacy and security emerge. Traditional topologies face some problems such as poor scalability, complicated management and inflexible communication. However, the huge cost of thousands of servers and switches is also a bottleneck for many companies to build or scale up data centers. This paper proposes a cost-efficient and flat topology, named Cell-Tree. It can achieve good performance and provide relatively high network capacity under a low cost. Accordingly, we designed a new communication mechanism and routing algorithm. The simulation shows that Cell-Tree has a relatively good performance with the cost significantly decreased.

Mesh-of-Torus: a new topology for server-centric data center networks

Various topologies have been proposed for high-performance computing (HPC), i.e., fat-tree, Torus topology. Compared with conventional fat-tree topology, Torus performs much better when applied in HPC. Unfortunately, due to its wraparound links, Torus topology naturally has the tendency to trigger deadlock incidents inside the network. Researchers solve this problem by means of virtual channel, but this approach will also restrict the routing of message. In this paper, we propose a deadlock-free topology for HPC, called Mesh-of-Torus, which incarnates the good characteristics of Mesh and Torus topology. Comparing with mesh, Mesh-of-Torus has shorter network diameter. Furthermore, we have proposed a corresponding port assignment rules in consideration of complicated internal arbitration or scheduling mechanism incurred by the employment of virtual channel. Deadlock avoidance can be achieved when dimension-order routing algorithm and our port assignment rules are applied to Mesh-of-Torus. Finally, simulations and mathematical analysis have shown that Mesh-of-Torus outperforms Mesh in terms of average end-to-end latency and network load distribution.