Data Center Network Topologies Research Articles

DINC: Toward Distributed In-Network Computing

In-network computing provides significant performance benefits, load reduction, and power savings. Still, an in-network service's functionality is strictly limited to a single hardware device. Research has focused on enabling on-device functionality, with limited consideration to distributed in-network computing. This paper explores the applicability of distributed computing to in-network computing. We present DINC, a framework enabling distributed in-network computing, generating deployment strategies, overcoming resource constraints and providing functionality guarantees across a network. It uses multi-objective optimization to provide a deployment strategy, slicing P4 programs accordingly. DINC was evaluated using seven different workloads on both data center and wide-area network topologies, demonstrating feasibility and scalability, providing efficient distribution plans within seconds.

Design and analysis of wireless data center network topology HCDCN based on VLC

With the rapid development of the Internet of Things, big data, cloud computing and other industries, the data center network, as the core infrastructure of the computing industry, is becoming increasingly important. Wireless communication technology is developing rapidly. This paper uses visible light wireless communication technology to build a wireless data center network, solving the problems of complex wiring and expansion difficulties in the existing data center network.Aiming at the problems of insufficient spectrum, low bandwidth utilization and large energy consumption in the existing wireless data center, this paper adopts the visible light wireless communication technology based on LED, and proposes a new wireless data center network structure HCDCN. Firstly, work have been done to analyze the feasibility of introducing data centers in this technology, and transform server racks in combination with the characteristics of the technology. Secondly, the construction process of the structure is introduced and a new coding rule is proposed. In order to increase the routing efficiency and fault tolerance, the rack top link routing path is added, and an efficient routing algorithm is designed. Finally, the aggregate throughput of HCDCN under different scales is obtained through experiments, and the network performance of HCDCN structure is compared with that of OWCells and Mesh structure. Experiments have proved that HCDCN's network performance is better than the other two structures under the same conditions. The new wireless data center network proposed in this paper has largely alleviated the bandwidth shortage of hotspot servers.

Study on Data Center Network Topologies for Monitoring Data using Edge/Fog Computing

The election of an appropriate data center network topology is key when dealing with surveillance and monitoring processes, such as those devoted to obtaining relevant data for sensor data fusion in any type of remote computing environment so as to perform fault diagnosis and prevention. In this paper, some of the most commonly used topologies to interconnect nodes within a data center bound to edge/fog computing, representing either switch-centric ones or server-centric ones, are reviewed and analyzed from a statistical point of view in order to measure their performance, resulting in server-centric ones doing it better.

Study of data center communication network topologies using complex network propagation model

Data center, as the core infrastructure of data storage and processing, calls for network security protection. Information security has been addressed in a number of specific ways. However, there are few studies that employ network topology features to prevent the transmission of viruses. When a virus spreads, different topologies display various properties. In this paper, we study three types of data center network topologies, i.e., Fat-tree, Leaf-spine, and Bcube, and quantify the propagation characteristics in every topology through the IC propagation model. The probability of the device being infected, the count of propagation sources, the access of propagation sources, and the topological parameters are all considered. Given that network security defenders can only change the topology and topological parameters, we propose a computational framework that combines factor analysis, which provides us with the selection of network topological parameters with a low virus propagation rate in the candidate parameter set. Through experiments, we find that Leaf-spine has a good inhibitory effect on viruses with high propagation probability. Meanwhile, each offers unique advantages. We hope that more data center network topologies will be studied to improve the security of all data centers using these network topologies.

Multicore fiber interconnects for multi-terabit spine-leaf datacenter network topologies

This work demonstrates the use of spatial division multiplexing (SDM) to implement spine-leaf datacenter network topologies. We consider two possible scenarios, with the first consisting of the direct use of high-capacity multicore fiber interconnects. This was implemented with an 8-core multicore fiber to support 64×200Gb/s SDM wavelength division multiplexing (WDM) lanes and combined throughput of 12.8 Tb/s. We also demonstrate a scenario in which each multicore fiber interconnect is used to establish multiple connections between pairs of spin-leaf switches. An optical circuit switch is used between the spine and leaf layers to establish core-selective connections, which significantly reduces the number of cables required to implement the network. We demonstrate this approach with a commercial optical circuit switch and 8-core and 4-core multicore fiber interconnects.

Performance Evaluation of Topologies for Multi-Domain Software-Defined Networking

Software-defined networking (SDN) is widely used in multiple types of data center networks, and these distributed data center networks can be integrated into a multi-domain SDN by utilizing multiple controllers. However, the network topology of each control domain of SDN will affect the performance of the multi-domain network, so performance evaluation is required before the deployment of the multi-domain SDN. Besides, there is a high cost to build real multi-domain SDN networks with different topologies, so it is necessary to use simulation testing methods to evaluate the topological performance of the multi-domain SDN network. As there is a lack of existing methods to construct a multi-domain SDN simulation network for the tool to evaluate the topological performance automatically, this paper proposes an automated multi-domain SDN topology performance evaluation framework, which supports multiple types of SDN network topologies in cooperating to construct a multi-domain SDN network. The framework integrates existing single-domain SDN simulation tools with network performance testing tools to realize automated performance evaluation of multi-domain SDN network topologies. We designed and implemented a Mininet-based simulation tool that can connect multiple controllers and run user-specified topologies in multiple SDN control domains to build and test multi-domain SDN networks faster. Then, we used the tool to perform performance tests on various data center network topologies in single-domain and multi-domain SDN simulation environments. Test results show that Space Shuffle has the most stable performance in a single-domain environment, and Fat-tree has the best performance in a multi-domain environment. Also, this tool has the characteristics of simplicity and stability, which can meet the needs of multi-domain SDN topology performance evaluation.

Fault-Tolerant Secure Routing Based on Trust Evaluation Model in Data Center Networks

With the rise of cloud computing technology, the whole information technology industry has undergone dramatic changes. Large-scale computing and the growing demand for computing power from large data have made it impossible for traditional data center network topology construction methods and network layer operation control methods with good stability to meet the requirements of technological development. The topology of a data center network is usually characterized as an undirected graph, and graph algorithms are used to implement communication patterns in complex network structures. However, the routing algorithms proposed earlier for graphs cannot be applied to all topological graphs. Based on the comprehensive interactive trust evaluation network security model, this study analyzes the advantages and disadvantages of relevant algorithms and proposes a fault-tolerant routing algorithm suitable for all networks. The algorithm can dynamically call different algorithms according to the number of current nodes and links. From the simulation results, it can be found that the algorithm not only improves the accuracy and fault tolerance but also effectively reduces the loss of time and memory.

Reliability assessment and profit analysis of distributed data center network topology

Reliability assessment and profit analysis of distributed data center network topology

A Novel Addressing and Routing Architecture for Cloud-Service Datacenter Networks

Datacenter networks (DCNs) play a key role in providing cloud services. The energy consumption and cost of a DCN are growing sharply with the extensions of network bandwidth and network size. The energy consumption, complexity and cost of a DCN depend on some design factors such as the topology structure, addressing scheme and routing mechanism. A novel addressing and routing architecture for cloud-service DCNs with regular topologies is proposed in this paper. First of all, we propose a port-based source-routing addressing (PSRA) scheme, which makes the table-lookup operation unnecessary and decreases the switch complexity. Next, leveraging the characteristics of PSRA and the regularity of DCN topologies, an extremely simple routing mechanism is designed, without switch involvement, control message interaction and topology information storage. Lastly, a high-efficiency fault-tolerance mechanism is proposed for the addressing and routing architecture. The analysis, implementation and simulation results indicate that the proposed architecture not only decreases the energy consumption and thus the cost of a DCN, but also enhances the routing performance and solves the fault-tolerance problem in a very efficient way.

CacheNet: Leveraging the principle of locality in reconfigurable network design

Emerging optical communication technologies support the dynamic reconfiguration of datacenter network topologies depending on the traffic they serve. However, to reap the benefits of such demand-aware networks, control logic that quickly learns and adapts to traffic patterns is required. This paper presents CacheNet, a novel approach to efficiently control demand-aware networks. CacheNet consists of two components, a demand-aware links-cache, and a demand-oblivious topology. CacheNet leverages temporal and spatial locality in the traffic by managing the reconfigurable links of the optical switches as a links-cache. Network traffic, in turn, can be served either by a link from the links-cache component or by a demand-oblivious topology component. We study several classic caching algorithms like online LFU and LRU as our caching algorithms, as well as offline optimal caching as a benchmark, and provide an analytical model which captures their performance benefits compared to an all demand-oblivious topology. Our analytical results show that based on the hit ratios and the links-cache size, when considering the average packet delay, our hybrid design outperforms a design that is based only on demand-oblivious topology. We also evaluate CacheNet empirically, using both synthetic and real-world traffic traces, confirming the potential of our approach to consider reconfigurable links as a network of links-cache.

An adaptive, fault tolerant, flow-level routing scheme for data center networks

Recent Data center network topologies put forward multiple paths between the end hosts to provide high bisection bandwidth. For the sake of utilizing the multiple paths efficiently many load-balancing techniques are proposed in literature. These load-balancing proposals mainly deal with distribution of traffic among multiple paths. Further, a good load-balancing technique scales down the latency of flows. While balancing the traffic among multiple paths, few proposals also take care of congestion. Flow-level solutions avoid flow splitting to prevent the packet re-ordering at the destination. Such adaptive solutions try to re-route (change the path) flows dynamically based on the feedback received from the network. In case of link failures, re-routing of the flow based on the queue size of the remote switch alone may be sub-optimal. The path to be chosen among multiple paths for re-routing, also poses a valid question. This paper proposes FlowFurl, an adaptive, distributed, end host driven and flow-level load-balancing technique. This adaptive routing technique supports dynamic flow-to-path assignment. This flow-to-path assignment is performed not only based on the congestion at the switch, but also the status of the links along the path. FlowFurl is the first technique to combine link failure and congestion information parameter for re-routing of flows. The proposed technique is deployment friendly, however it incurs small amount of overhead (small amount of memory at each switch to maintain the link state). We show that our proposed scheme FlowFurl performs better than considered schemes in terms of average flow latency, percentage of flows that achieve deadline and average flow throughput, by utilizing local information such as connectivity information with congestion information.

Performance Analysis of a Graph-Theoretic Load Balancing Method for Data Centers

Modern data centers can process a massive amount of data in a short time with minimal errors. Data center networks (DCNs) use equal-cost, multi-path topologies to deliver split flows across alternative paths between the core layer and hosted servers, which could lead to significant overload if path scheduling is inefficient. Thus, distributing incoming requests among these paths is crucial for providing higher throughput and protection against link or switch failures. Several approaches have been proposed for path selection, mainly relying on round-robin and least-congested methods. In this paper, we propose a load-balancing method based on betweenness centrality to improve the overall performance of a data center in terms of throughput, delay, and energy consumption. For evaluation, we compare our method with baseline methods of different DCN topologies: fat-tree, DCell, and BCube. On average, the evaluation results show that our method outperforms the others. It increases throughput by 202% and 33% while reducing delay by 20% and 22%, and energy consumption by 40% and 41% compared to the round-robin and least-congested methods, respectively.

CRetor: An SDN-Based Routing Scheme for Data Centers With Regular Topologies

The blooming of cloud computing leads to the rapid expansion of data center networks (DCN). Conventional link state routing algorithms like OSPF are widely adopted in data center networks, however, those routing algorithms bring great control overheads and long convergence time. Recently, topology-aware routing methods are considered to be efficient especially in data center networks with regular topologies. Lots of topology-aware routing methods (e.g., Fat-Tree and BCube) have been proposed for specific data center network topologies. This paper first proposes a formalized method to describe regular topologies and a regular Topology Description Language (TPDL) based on this method. TPDL is well designed to accurately define regular network topologies in a clear way leveraging their regularities. Based on the Software-Defined Networking (SDN) technology, this paper also proposes a novel topology-aware routing scheme: cRetor (controller-side REgular TOpology Routing scheme). Different from other topology-aware routing methods, cRetor is a TPDL-based general routing method, which means it is expected to work on different kinds of regular topologies. In this scheme, TDPL files are used as a priori knowledge to build an initial topology in the SDN controllers, which eliminates the process of topology discovery via Link Layer Discovery Protocol (LLDP) and hence relieves the bandwidth and processing burdens on controllers. Besides, we also apply the A-star algorithm to SDN controllers to speed up the routing selection, where TPDL's distance formulas act as the heuristic function. The experimental results show that cRetor outperforms LLDP-based SDN, OSPF and DCell in routing calculation performance, convergence speed, routing overheads and fault tolerance.

An Ant Colony Optimization Based Data Update Scheme for Distributed Erasure-Coded Storage Systems

Owing to the high availability and space-efficiency of erasure codes, they have become the de facto standard to provide data durability in large scale distributed storage systems. The update-intensive workloads of erasure codes lead to a large amount of data transmission and I/O. As a result, it becomes a major challenge to reduce the amount of data transmission and optimize the use of existing network resources so that the update efficiency of the erasure codes could be improved. However, very little research has been done to optimize the update efficiency of the erasure codes under multiple QoS metrics. In this paper, our proposed update scheme, the Ant Colony Optimization based multiple data nodes Update Scheme (ACOUS) employs a two-stage rendezvous data update procedure to optimize the multiple data nodes updates. Specifically, the two-stage rendezvous data update procedure performs the data delta collection and the parity delta distribution based on a multi-objective update tree which is built by the ant colony optimization routing algorithm. Under typical data center network topologies, extensive experimental results show that, compared to the traditional TA-Update scheme, our scheme is able to achieve a 26% to 37% reduction of update delay with convergence guarantee at the cost of negligible computation overhead.

Joint virtual middlebox and application placement with bandwidth guarantees in multi-tenant datacenter

Hardware middleboxes are widely used in datacenter to provide network functions such as firewalls, intrusion detection system, load balancers, etc. Unfortunately, the hardware middleboxes are unable to offer customized functions for individual tenant. To this end, the software middleboxes have been introduced in datacenter which provide flexible network functions for customers. The arrival traffic of one tenant is first distributed to the pre-allocated software middleboxes chain and then delivered to the application servers of the tenant. With regard to the complicated communication of intra-middleboxs, middleboxes-to-applications and intra-applications traffics, we propose vMAP, a joint middlebox and application placement algorithm to effectively use the limited bandwidth inside datacenter. vMAP leverages the communication model of tenant’s requirement for both middlebox and application and provides an effective virtual machines (VMs) placement algorithm to minimize the bandwidth consumption inside datacenters. vMAP further addresses the middlebox scaling problem by reserving few free VMs resource based on a probability model of the variable workload. Extensive simulations on current large-scale datacenter network topologies verify that vMAP can accept more tenant requests than prior placement algorithm (e.g., Stratos (Gember et al. in Stratos: Virtual middleboxes as first-class entities, University of Wisconsin-Madison, Madison, 2012) ) while saving $$\sim$$10% core-level bandwidth.

Topology-aware multi-objective virtual machine dynamic consolidation for cloud datacenter

With help of virtualization technology, cloud datacenters provide affordable computation power to numerous enterprises and business users. Virtual machine dynamic consolidation (VMDC) plays an important role in the energy consumption management of datacenters. Previous solutions on VMDC either fail to consider impact resulting from the interaction between datacenter network topology and communication topology, or handle this issue in coarse granular level. In this paper, we propose a new approach to coordinate datacenter network topology and communication topology. The proposed VMDC approach adopts a multi-objective genetic algorithm to simultaneously minimize energy consumption and communication traffic causing performance bottlenecks, while meeting SLA constraint on migration time. In this approach, both SLA constraint and objective function on communication traffic are formulated with topology awareness in switch link level. A comprehensive simulation study shows that the proposed approach can significantly save energy consumption of datacenters and provide SLA assurance for different types of datacenter topologies. Simulation results demonstrate that the proposed approach outperforms prior approaches in terms of reducing energy consumption, bottleneck ratio and SLA violation ratio.

Data center network energy consumption minimization: a hierarchical FAT-tree approach

To meet the ever-growing cloud users’ demands, deployment of the large-scale datacenters has taken place and the numbers of servers have been increased exponentially. In modern years, there have been hasty growth of geographically distributed and large-scale data centers which is been deployed to support current users’ demands by the internet service operator. These datacenters consume enormous amount of electric power that includes servers, cooling-machines and networking switches, which leads to enormous challenges in the era of designing cost-effective and efficient data center network. In recent year, number of works proposed to overcome the above challenges in different direction. In this paper, different datacenter network-topologies have been investigated and a modified cost effective hierarchical Fat-Tree for building network-topology has been proposed. After this, we articulate the energy optimization problem as a MILP problem with an objective to minimize the amount of energy consumed in datacenter networks. Due to effective high computational complexity of MILP, an algorithm is proposed to deal with the problem of Several extensive simulation-based experiments are accompanied using an optimization solver (CPLEX) to evaluate and validate the effectiveness of the given proposed algorithm. The efficiency of the proposed model is shown by the several simulation-based results.

Global Round Robin: Efficient Routing With Cut-Through Switching in Fat-Tree Data Center Networks

Fat tree is a scalable and widely deployed data center network topology. In this paper, a novel framework for designing per-packet load-balanced routing algorithms in fat tree called global round robin (GRR) is proposed. Routing in fat tree consists of uprouting and downrouting. In uprouting, a packet is sent to a switch that is a common ancestor (CA) of the source (server) and the destination. In downrouting, the packet is sent from the CA switch to the destination. Assume that time is slotted and each slot can accommodate one packet. With GRR, in each slot, a connection configuration is formed by establishing an uprouting path from each server to a spine switch port such that no paths will cross each other. Packets are sent from sources to respective spine switches with cut-through switching. The connection configuration is updated in a round robin fashion such that in every $m$ slots, where $m$ is the number of spine switches, each server is connected to each spine switch exactly once. Since a CA does not need to be a spine switch, an improved GRR (IGRR) is then proposed to allow the nearest CA to intercept packets for downrouting. We prove that both GRR and IGRR can guarantee 100% throughput under a wide class of traffic. An analytical model is also constructed for studying their delay performance under uniform traffic. Finally, simulation results show that IGRR provides the best delay-throughput performance among all the existing per-packet load-balanced routing algorithms.

A New Approach to Stochastic Scheduling in Data Center Networks

The Quality of Service (QoS) of scheduling between latency-sensitive small data flows (a.k.a. mice) and throughput-oriented large ones (a.k.a. elephants) has become ever challenging with the proliferation of cloud-based applications. In light of this mounting problem, this work proposes a novel flow control scheme, HOLMES (HOListic Mice-Elephants Stochastic), which offers a holistic view of global congestion awareness as well as a stochastic scheduler of mixed mice-elephants data flows in Data Center Networks (DCNs). Firstly, we theoretically prove the necessity for partitioning DCN paths into sub-networks using a stochastic model. Secondly, the HOLMES architecture is proposed, which adaptively partitions the available DCN paths into low-latency and high-throughput sub-networks via a global congestion-aware scheduling mechanism. Based on the stochastic power-of-two-choices policy, the HOLMES scheduling mechanism acquires only a subset of the global congestion information, while achieves close to optimal load balance on each end-to-end DCN path. We also formally prove the stability of HOLMES flow scheduling algorithm. Thirdly, extensive simulation validates the effectiveness and dependability of HOLMES with select DCN topologies. The proposal has been in test in an industrial production environment. An extensive survey of related work is also presented.

Centralized approaches for virtual network function placement in SDN-enabled networks

Software-defined networking (SDN) and network function virtualization (NFV) represent significant changes to the architecture of data networks. SDN provides clean separation of the control plane from the data plane while NFV helps virtualize functions typically implemented using middleboxes into virtual network functions (VNFs). The network function placement (NFP) problem involves placing VNFs to satisfy the service function chaining (SFC) requests of the flows in the network. Current solutions to this problem are slow and cannot handle real-time requests. In this paper, a static NFP solution that uses a divide-and-conquer approach is discussed first, with complexity similar to that of existing solutions. It is shown that the solution is complete and sound. Next, customization of this solution to obtain an agile version that trades off precision for time complexity is discussed. A combination of this divide-and-conquer solution with a modified version of Dijkstra’s algorithm is used to solve the problem for dynamic requests. Finally, a centralized, online SDN-enabled approach to solving the problem is proposed. The proposed architecture and algorithms are simulated and analyzed with various system parameters and are shown to scale to large data center network (DCN) topologies.