Data Center Network Topologies Research Articles

Striking a Balance Between Traffic Engineering and Energy Efficiency in Virtual Machine Placement

The increasing adoption of server virtualization has recently favored three key technology advances in data-center networking: the emergence at the hypervisor software level of virtual bridging functions between virtual machines and the physical network; the possibility to dynamically migrate virtual machines across virtualization servers in the data-center network (DCN); a more efficient exploitation of the large path diversity by means of multipath forwarding protocols. In this paper, we investigate the impact of these novel features in DCN optimization by providing a comprehensive mathematical formulation and a repeated matching heuristic for its resolution. We show, in particular, how virtual bridging and multipath forwarding impact common DCN optimization goals, traffic engineering (TE) and energy efficiency (EE), and assess their utility in the various cases of four different DCN topologies. We show that virtual bridging brings a high performance gain when TE is the primary goal and should be deactivated when EE becomes important. Moreover, we show that multipath forwarding can bring relevant gains only when EE is the primary goal and virtual bridging is not enabled.

HERO: Hierarchical Energy Optimization for Data Center Networks

The rapid escalating power consumption has become critically important to modern data centers. Existing works on reducing the power consumption of network elements formulate the power optimization problem for a general network topology and require a centralized controller. As the scale of data centers increases, the complexity of solving this optimization problem increases rapidly. Inspired from the hierarchical data center network (DCN) topologies and data center traffic patterns, we design a two-level, pod-level, and core-level power optimization model, namely, Hierarchical EneRgy Optimization (HERO), to reduce the power consumption of network elements by switching off network switches and links while still guaranteeing full connectivity and maximizing link utilization. Given a physical DCN topology and a traffic matrix, we illustrate that two-level power optimizations in HERO fall in the class of capacitated multicommodity minimum cost flow (CMCF) problem, which is NP-hard. Therefore, we design several heuristic algorithms based on different switch elimination criteria to solve the proposed HERO optimization problem. The power-saving performance of the proposed HERO model is evaluated by several experiments with different traffic patterns. Our simulations demonstrate that HERO can reduce power consumptions of network elements effectively with reduced complexity.

Network Element Scheduling for Achieving Energy-Aware Data Center Networks

The goal of data center network is to interconnect a massive number of servers so as to provide reliable and scalable computing and storage infrastructure for cloud-based Internet services and data-intensive scientific applications. Recent studies reveal that the network elements consume 10∼20% of the overall power in a data center, which has introduced a challenge to reducing network energy cost without adversely affecting network performance. Considering unique features of traffic patterns and network topologies in data centers, this paper proposes a novel Network Element Scheduling Scheme (NESS) to reduce data center energy consumption from the networking perspective. The core idea is to turn on only a minimal subset of network elements to satisfy routing requirements, and put to sleep or shut down the rest unneeded ones for energy saving. In NESS, the logical network architecture formed by the active elements not only achieves the basic purpose for server interconnections in data centers, but also can support multi-path routing between pairs of hot servers for load balancing. Simulation experiments are performed in representative data center network topologies, and the results demonstrate the effectiveness of NESS in energy conserving on network elements in data centers.

Twin Datacenter Interconnection Topology

The authors propose an interconnection topology suited to server-centric networks based on Twin graphs. The topology exploits graph theory to cost-effectively improve network scalability and resilience. The authors discuss datacenter network topology by comparing Twin interconnection topologies to different datacenter topologies (that is, Fat Tree, BCube, DCell, and Hypercube). Their results show that Twin topologies have a lower link cost than previous datacenter topologies, are scalable with fine granularity, and are efficient, fault tolerant, and resilient owing to their features, which enable traffic balancing under both normal and faulty operating conditions.

Decluster: a complex network model-based data center network topology

To cope with increasing demands of computation and storage, data centers should follow the pace of the rapid growth of data size. It is necessary for a data center with a scalability property of which each expansion of a data center network is done with a few modifications. Besides the scalability property, we also need a data center to have good performance, such as high throughput. For these purposes, we propose Decluster, a complex network model-based data center network topology. The complex network model of Decluster is derived from a random network. Such a model just satisfies the requirement of scalability. Decluster employs a complex network model to achieve high throughput via reducing the variance of local clustering coefficients. We have carried out extensive simulations to demonstrate that Decluster enjoys good performance while keeping scalability.

Measuring throughput of data center network topologies

High throughput is a fundamental goal of network design. While myriad network topologies have been proposed to meet this goal, particularly in data center and HPC networking, a consistent and accurate method of evaluating a design's throughput performance and comparing it to past proposals is conspicuously absent. In this work, we develop a framework to benchmark the throughput of network topologies and apply this methodology to reveal insights about network structure. We show that despite being commonly used, cut-based metrics such as bisection bandwidth are the wrong metrics: they yield incorrect conclusions about the throughput performance of networks. We therefore measure flow-based throughput directly and show how to evaluate topologies with nearly-worst-case traffic matrices. We use the flow-based throughput metric to compare the throughput performance of a variety of computer networks. We have made our evaluation framework freely available to facilitate future work on design and evaluation of networks.

The importance of switch dimension for energy-efficient datacenter design

Saving power in datacenter networks has become a pressing issue as the networks tend to consume steady power even when many servers may be idle during periods of low activity. While in operation, ElasticTree and CARPO can save power consumed by a fat-tree network by using sleep mode, where some components such as ports and switches are turned off when traffic demand in the network is relatively moderate. In this paper, we propose a new approach by exploring the design stage of a datacenter network, and focus on how to choose the right switch size that can potentially save the most power during the expected operation of the network. We also consider speed scaling where the power of a switch can be varied by adjusting its processing rate according to its traffic demand. We first perform analysis and simulation to investigate the power-saving performance of different switch sizes, power-saving modes and traffic demand patterns. Based on fat-tree networks with sleep mode and supporting a fixed number of servers, our findings reveal that deploying a large number of small switches is more power-efficient than a small number of large switches when the traffic demand is relatively moderate or when servers exchanging traffic are in close proximity. With speed scaling, the reverse is generally true. We confirm our findings using traffic traces from a production datacenter. Our approach for studying the fat-tree topology design is also extendible to other types of datacenter network topologies.

Datacast: A Scalable and Efficient Reliable Group Data Delivery Service for Data Centers

Reliable Group Data Delivery (RGDD) is a pervasive traffic pattern in data centers. In an RGDD group, a sender needs to reliably deliver a copy of data to all the receivers. Existing solutions either do not scale due to the large number of RGDD groups (e.g., IP multicast) or cannot efficiently use network bandwidth (e.g., end-host overlays). Motivated by recent advances on data center network topology designs (multiple edge-disjoint Steiner trees for RGDD) and innovations on network devices (practical in-network packet caching), we propose Datacast for RGDD. Datacast explores two design spaces: 1) Datacast uses multiple edge-disjoint Steiner trees for data delivery acceleration. 2) Datacast leverages in-network packet caching and introduces a simple soft-state based congestion control algorithm to address the scalability and efficiency issues of RGDD. Our analysis reveals that Datacast congestion control works well with small cache sizes (e.g., 125KB) and causes few duplicate data transmissions (e.g., 1.19%). Both simulations and experiments confirm our theoretical analysis. We also use experiments to compare the performance of Datacast and BitTorrent. In a BCube(4, 1) with 1Gbps links, we use both Datacast and BitTorrent to transmit 4GB data. The link stress of Datacast is 1.01, while it is 1.39 for BitTorrent. By using two Steiner trees, Datacast finishes the transmission in 16.9s, while BitTorrent uses 52s.

ZUpdate

Datacenter networks (DCNs) are constantly evolving due to various updates such as switch upgrades and VM migrations. Each update must be carefully planned and executed in order to avoid disrupting many of the mission-critical, interactive applications hosted in DCNs. The key challenge arises from the inherent difficulty in synchronizing the changes to many devices, which may result in unforeseen transient link load spikes or even congestions. We present one primitive, zUpdate, to perform congestion-free network updates under asynchronous switch and traffic matrix changes. We formulate the update problem using a network model and apply our model to a variety of representative update scenarios in DCNs. We develop novel techniques to handle several practical challenges in realizing zUpdate as well as implement the zUpdate prototype on OpenFlow switches and deploy it on a testbed that resembles real DCN topology. Our results, from both real-world experiments and large-scale trace-driven simulations, show that zUpdate can effectively perform congestion-free updates in production DCNs.

DAC: Generic and Automatic Address Configuration for Data Center Networks

Data center networks encode locality and topology information into their server and switch addresses for performance and routing purposes. For this reason, the traditional address configuration protocols such as DHCP require a huge amount of manual input, leaving them error-prone. In this paper, we present DAC, a generic and automatic Data center Address Configuration system. With an automatically generated blueprint that defines the connections of servers and switches labeled by logical IDs, e.g., IP addresses, DAC first learns the physical topology labeled by device IDs, e.g., MAC addresses. Then, at the core of DAC is its device-to-logical ID mapping and malfunction detection. DAC makes an innovation in abstracting the device-to-logical ID mapping to the graph isomorphism problem and solves it with low time complexity by leveraging the attributes of data center network topologies. Its malfunction detection scheme detects errors such as device and link failures and miswirings, including the most difficult case where miswirings do not cause any node degree change. We have evaluated DAC via simulation, implementation, and experiments. Our simulation results show that DAC can accurately find all the hardest-to-detect malfunctions and can autoconfigure a large data center with 3.8 million devices in 46 s. In our implementation, we successfully autoconfigure a small 64-server BCube network within 300 ms and show that DAC is a viable solution for data center autoconfiguration.

A Novel Method of Fault-Tolerant Decentralized Lookup Service for the Cloud Computing

With the rapid development of cloud computing, data center network scale exponentially, and the traditional tree network topology have not accommodate this growth model for a single point of failure and bottlenecks. DCell is a decentralized, fault-tolerant data center network topology. Although DCell has good fault tolerance, it does not implement the strategy of the data object placemen and query. This paper proposes a novel model NetCloud of data placement and query for cloud computing in DCN based on DCell. In the NetCloud, the key of data item is be mapped to a binary string by the NetCloud_hash algorithms. We present placement and the fast locating strategy of the data object in the NetCloud. We analyze performance of NetCloud. The simulation shows that it has better query performance and fault tolerance.

Generic and automatic address configuration for data center networks

Data center networks encode locality and topology information into their server and switch addresses for performance and routing purposes. For this reason, the traditional address configuration protocols such as DHCP require huge amount of manual input, leaving them error-prone. In this paper, we present DAC, a generic and automatic Data center Address Configuration system. With an automatically generated blueprint which defines the connections of servers and switches labeled by logical IDs, e.g., IP addresses, DAC first learns the physical topology labeled by device IDs, e.g., MAC addresses. Then at the core of DAC is its device-to-logical ID mapping and malfunction detection. DAC makes an innovation in abstracting the device-to-logical ID mapping to the graph isomorphism problem, and solves it with low time-complexity by leveraging the attributes of data center network topologies. Its malfunction detection scheme detects errors such as device and link failures and miswirings, including the most difficult case where miswirings do not cause any node degree change. We have evaluated DAC via simulation, implementation and experiments. Our simulation results show that DAC can accurately find all the hardest-to-detect malfunctions and can autoconfigure a large data center with 3.8 million devices in 46 seconds. In our implementation, we successfully autoconfigure a small 64-server BCube network within 300 milliseconds and show that DAC is a viable solution for data center autoconfiguration.

Risk-neutral evaluation of information security investment on data centers

Based on given data center network topology and risk-neutral management, this work proposes a simple but efficient probability-based model to calculate the probability of insecurity of each protected resource and the optimal investment on each security protection device when a data center is under security breach. We present two algorithms that calculate the probability of threat and the optimal investment for data center security respectively. Based on the insecurity flow model (Moskowitz and Kang 1997) of analyzing security violations, we first model data center topology using two basic components, namely resources and filters, where resources represent the protected resources and filters represent the security protection devices. Four basic patterns are then identified as the building blocks for the first algorithm, called Accumulative Probability of Insecurity, to calculate the accumulative probability of realized threat (insecurity) on each resource. To calculate the optimal security investment, a risk-neutral based algorithm, called Optimal Security Investment, which maximizes the total expected net benefit is then proposed. Numerical simulations show that the proposed approach coincides with Gordon's (Gordon and Loeb, ACM Transactions on Information and Systems Security 5(4):438---457, 2002) single-system analytical model. In addition, numerical results on two common data center topologies are analyzed and compared to demonstrate the effectiveness of the proposed approach. The technique proposed here can be used to facilitate the analysis and design of more secured data centers.