Inter-datacenter Networks Research Articles

Adaptability Analysis for IP Switching and Optical Switching in Geographically Distributed Inter-Datacenter Networks

The unexpected growth of cloud services lead to a dramatic increase in the network capital expenditure (CAPEX) and latency in geographically distributed inter-datacenter networks. In this paper, an adaptability analysis is proposed on CAPEX and forwarding delay for IP switching and optical transport network (OTN) switching in distributed inter-datacenter networks. By abstracting the switching behavior of the IP routers and OTN equipment, the IP nodes, and OTN nodes are mapped to the switching matrix of CLOS switching network. A unified network-level connection model is established based on CLOS switching network to meet the needs of connectivity. Then, the hardware costs and the forwarding delay are evaluated to find out the bandwidth threshold (BDT) which decides the most suitable switching mode for a specific switching granularity. The adaptability of IP switching and OTN switching under different switching granularities and port rates is performed based on the simulation platform in terms of hardware cost, optical port cost, and forwarding delay. Simulation results show that the BDT based on hardware cost is 280 Mbps when switching granularity is 1.25 Gb/s. For 2.5 Gb/s switching granularity, when average bandwidth between multiple datacenters is higher than 600 Mbps, OTN switching has more advantages than IP switching on capital expenditure, especially when port rate of the equipment is high. In addition, the simulation results also verify that the forwarding delay is reduced obviously with OTN switching.

SDN Enabled Restoration With Triggered Precomputation in Elastic Optical Inter-Datacenter Networks

Flexi-grid elastic optical networks (EONs) provide a promising infrastructure for meeting the high-bandwidth and low-latency requirements of interconnecting datacenters. Since failures of high-capacity fiber links prevent users from accessing cloud services and lead to large amounts of data loss, much recent research has focused on inter-datacenter network failure recovery. In this paper, we build upon software-defined networking (SDN) technology and develop a fast restoration strategy with triggered precomputation (FR-TP) that achieves fast failure recovery with minimal resource overhead. Our work makes several contributions. We extend the controller functionality and OpenFlow protocol for software-defined elastic optical inter-datacenter network architectures, so as to obtain the global network state information quickly and accurately. We then present the FR-TP strategy and an associated trigger mechanism to compute backup paths before a link failure occurs. Furthermore, to improve the efficiency of path computation and bandwidth resource assignment, we construct a layered auxiliary graph of spectrum window planes (SWP-LAG) using a residual capacity matrix to change dynamically the width of the spectrum window planes (SWPs) to satisfy different service requests. Simulation results demonstrate that, compared with existing restoration strategies, the proposed FR-TP strategy combined with SWP-LAG reduces the recovery time by up to 30.4% in the network topologies studied without increasing the blocking probability.

Manycast, Anycast, and Replica Placement in Optical Inter-Datacenter Networks

The expanding adoption of cloud-based services in recent years puts stringent requirements on datacenters (DCs) and their interconnection networks. Optical inter-datacenter networks represent the only viable option for satisfying the huge bandwidth required to replicate and update content for cloud-based services across geographically dispersed datacenters. In addition to content replication and synchronization, optical inter-datacenter networks must also support communication between datacenters and end-users. The resulting new traffic patterns and the enormous traffic volumes call for new capacityefficient approaches for inter-datacenter network designs that incorporate both transport and datacenter resource planning. This paper introduces an integrated approach to optimally place content replicas across DCs by concurrently solving the routing and wavelength assignment (RWA) problem for both inter-DC content replication and synchronization traffic following the manycast routing paradigm, and end-user-driven user-to-DC communication following the anycast routing paradigm, with the objective to reduce the overall network capacity usage. To attain this goal, the manycast, anycast, and replica placement (MARP) problem is formulated as an integer linear program to find optimal solutions for smaller problem instances. Due to the problem complexity, a scalable and efficient heuristic algorithm is developed to solve larger network scenarios. Simulation results demonstrate that the proposed integrated MARP strategy can significantly reduce the network capacity usage when compared to the benchmarking replica placement and RWA schemes aimed at minimizing the resources consumed by either of the two types of traffic independently.

A spectrum-efficient algorithm based on traffic splitting and merging transmission for anycast in inter-datacenter elastic optical networks

Anycast is attracting much attention due to the need of scalable and cost-effective data delivery in inter-datacenter elastic optical networks. However, spectrum fragmentation degrades network’s performance and decreases probability of successful anycast delivery significantly. When the idle spectrum block in elastic optical network is not enough to transmit the anycast request, spectrum splitting with multi-path transmitting the anycast is an effective approach to improve the spectrum fragmentation utilization. For improving spectrum utilization and reducing time delay between multiple paths, we propose a spectrum-efficient algorithm based on traffic splitting and merging (Anycast_SA_TSM) transmission to avoid spectrum fragmentation and delay between multiple paths. In order to minimize the time delay between multiple paths, we design a modified scheme to select the multiple paths with minimal time delay to transmit the anycast. During the spectrum allocation phase, a new spectrum block allocation scheme, the exact fit or fragmentation minimal, is put forward. Moreover, when an appropriate size of spectrum block is found for the split anycast request, merging split sub-requests to a single path is activated for minimizing the additional guard bands and improving the spectrum efficiency. Comparing with other two anycast algorithms, simulation results show that the proposed algorithm can get minimal time delay between split multiple paths, the minimal bandwidth blocking probability and the highest spectrum utilization.

<italic>Stemflow:</italic> Software-Defined Inter-Datacenter Overlay as a Service

Modern Internet applications are typically hosted in the public cloud, with multiple server instances running within geographically distributed datacenters. Thanks to the abundantly available bandwidth on wide-area links that interconnect these datacenters, it is conceivable that bandwidth-intensive applications may improve their performance by relaying their traffic through such an inter-datacenter network . However, there does not yet exist a cloud service that provides a turn-key solution to tap into such available bandwidth resources conveniently. In this paper, we design and implement Stemflow , a new system framework that provides Inter-Datacenter Overlay as a Service based on the software-defined networking principle. It offers an attractive foundation that helps an Internet application to transparently improve its scalability and performance by using inter-datacenter networks for its traffic. With Stemflow , all deployed server instances will construct an overlay atop an inter-datacenter network, and the routing decisions to relay application traffic are made by a centralized controller. The algorithms needed to make these decisions are customized to meet the needs of individual applications, and are cached within the data plane. We motivate and describe the design decisions, and present an extensive experimental evaluation in public cloud infrastructures, using two example applications as our case studies.

Crosstalk-aware virtual network embedding over inter-datacenter optical networks with few-mode fibers

Abstract Virtualization of datacenter (DC) infrastructures enables infrastructure providers (InPs) to provide novel services like virtual networks (VNs). Furthermore, optical networks have been employed to connect the metro-scale geographically distributed DCs. The synergistic virtualization of the DC infrastructures and optical networks enables the efficient VN service over inter-DC optical networks (inter-DCONs). While the capacity of the used standard single-mode fiber (SSMF) is limited by their nonlinear characteristics. Thus, mode-division multiplexing (MDM) technology based on few-mode fibers (FMFs) could be employed to increase the capacity of optical networks. Whereas, modal crosstalk (XT) introduced by optical fibers and components deployed in the MDM optical networks impacts the performance of VN embedding (VNE) over inter-DCONs with FMFs. In this paper, we propose a XT-aware VNE mechanism over inter-DCONs with FMFs. The impact of XT is considered throughout the VNE procedures. The simulation results show that the proposed XT-aware VNE can achieves better performances of blocking probability and spectrum utilization compared to conventional VNE mechanisms.

Traffic Allocation Strategies in WSS-Based Dynamic Optical Networks

Elastic optical networking (EON) is a viable solution to meet future dynamic capacity requirements of Internet service provider and inter-datacenter networks. At the core of EON, wavelength selective switches (WSSs) are applied to individually route optical circuits, while assigning an arbitrary bandwidth to each circuit. Critically, the WSS control scheme and configuration time may delay the creation time of each circuit in the network. In this paper, we first detail the WSS-based optical data-plane implementation of a metropolitan network test-bed. Then, we review a software-defined networking (SDN) application designed to enable dynamic and fast circuit setup. Subsequently, we introduce a WSS logical model that captures the WSS time-sequence and is used to estimate the circuit-setup response time. Then, we present two batch service policies that aim to reduce the circuit-setup response time by bundling multiple WSS reconfiguration steps into a single SDN command. Resulting performance gains are estimated through simulation.

On the performance of the wide-area networks interconnecting public-cloud datacenters around the globe

According to current usage patterns, research trends, and latest reports, the performance of the wide-area networks interconnecting geographically distributed cloud nodes (i.e. inter-datacenter networks) is gaining more and more interest. In this paper we leverage only active approaches—thus we do not rely on information restricted to providers—and propose a deep analysis of these infrastructures for the two public-cloud leading providers: Amazon Web Services and Microsoft Azure. Our study provides an assessment of the performance of these networks as a function of the several configuration factors under the control of the customer and evidences specific cases of particular interest. The analysis of these cases and of their root causes, also related with service fees, provides insights on their impact on both the Quality of Service perceived by cloud customers and the outcomes of studies neglecting them.Our results show that Azure inter-datacenter infrastructure performs better than Amazon’s in terms of throughput (+56%, on average). On the other hand, the performance of the two providers is comparable in terms of latency, with the exception of limited specific cases. Moreover, some of the configuration factors cloud customers can leverage (such as larger more expensive VM sizes, advertised to have better network performance) may have no effect on the inter-datacenter network performance actually perceived. Counterintuitively, lower performance may even be related to higher costs for the customer. Experimental evidences show that public-cloud providers also rely on external network providers for some geographical regions, which is the cause of lower performance and higher costs. A comparison with previous works show that TCP throughput has not been improved recently, while evidences of higher link capacities have been found.

Inter-Data Center Network Dimensioning under Time-of-Use Pricing

In the cloud era, data centers consume tremendous power due to their huge computing and storage requirements. Furthermore, allocation and release of resources by numerous cloud customers leads to significant energy consumption at the data centers, which in turn, increases the Operational expenditures (Opex) of the operators. In this article, we combine energy efficiency and Time of Use (ToU)-awareness, and propose a novel virtualization scheme, namely ToU-aware Provisioning (ToUP) for an inter-data center network over an IP over WDM backbone. In ToUP, in addition to the traffic between two backbone nodes, upstream user demands destined to data centers and downstream data center demands originating from many data centers; inter-data center traffic is also considered for workload sharing between the data centers. Initially, we present an MILP formulation to model the optimal behavior of ToUP. Since the inter-data center network needs to be reconfigured in polynomial time, we propose a simulated annealing (SA)-based heuristic. We verify the heuristic by using the MILP solution as the benchmark. We evaluate ToUP under various scenarios, and numerical results confirm that significant Opex savings can be achieved while demands can be provisioned with low energy consumption in the data centers and network equipments.

Resiliency versus energy sustainability in optical inter-datacenter networks

With cloud computing becoming a forefront of business models, saving in operational expenses is one of the main challenges and goals of the operators. The expenses for the operators come directly from energy consumption, and indirectly through service downtime i.e. losing customers or paying penalties. This paper considers both energy efficiency and resiliency design under different topological characteristics where datacenters are inter-connected via an elastic optical network backbone. We present a benchmark scheme which aims at minimum outage probability provisioning, and a compromising scheme which aims at resilient provisioning with minimum power consumption. The former is solely concerned with minimizing outages while the latter combines energy awareness while minimizing outages. Through simulations, we show that introducing energy awareness into outage minimized provisioning reduces the power consumption by about 7%. When compared against energy-aware provisioning, the compromising scheme can reduce outage probabilities by 45–97%. The compromising approach even improves upon the other algorithms by introducing shorter path delays. We further investigate the impact of electric price-aware and resilient inter-datacenter network design in the presence of dynamic electricity markets. We show that the electric bills of network operators can be cut up to 50% if Time of Use-awareness is adopted.

Orchestrating Tree-Type VNF Forwarding Graphs in Inter-DC Elastic Optical Networks

It is known that by incorporating network function virtualization (NFV) in inter-datacenter (inter-DC) networks, service providers can use their network resources more efficiently and adaptively and expedite the deployment of new services. This paper studies the provisioning algorithms to realize tree-type virtual network function forwarding graphs (VNF-FGs), i.e., multicast NFV trees (M-NFV-Ts), in inter-DC elastic optical networks (IDC-EONs) cost-effectively. Specifically, we try to optimize the VNF placement and multicast routing and spectrum assignment jointly for orchestrating M-NFV-Ts in an IDC-EON with the lowest cost. Our study addresses both static network planning and dynamic network provisioning. For network planning, we first formulate a mixed integer linear programming (MILP) model to solve the problem exactly, and then propose three heuristic algorithms, namely, auxiliary frequency slot matrix (AFM)-MILP, AFM-GS, and RB. Extensive simulations show that AFM-MILP and AFM-GS can approximate the MILP's performance on low-cost M-NFV-T provisioning with much shorter running time. For network provisioning, we design two additional online algorithms based on AFM-GS and RB to serve M-NFV-Ts in a dynamic IDC-EON, with the consideration of spectrum fragmentation.

Maximizing Utility of Time-Constrained Emergency Backup in Inter-Datacenter Networks

This work addresses the emergency backup problem in inter-datacenter (inter-DC) network, which is triggered in response to predictable destructive events. We consider the value of data and study how to maximize the utility of time-constrained emergency backup. Specifically, we first formulate the utility-maximization problem for a time-varying network environment and then leverage the time-expanded network (TEN) approach to transform it into a static network problem. To finally solve the problem, a time-efficient and distributed algorithm is proposed based on the dual decomposition technique. Simulation results show that our algorithm outperforms existing algorithms.

Delay-Optimized Video Traffic Routing in Software-Defined Interdatacenter Networks

Many video streaming applications operate their geo-distributed services in the cloud, taking advantage of superior connectivities between datacenters to push content closer to users or to relay live video traffic between end users at a higher throughput. In the meantime, inter-datacenter networks also carry high volumes of other types of traffic, including service replication and data backups, e.g., for storage and email services. It is an important research topic to optimally engineer and schedule inter-datacenter traffic, taking into account the stringent latency requirements of video flows when transmitted along inter-datacenter links shared with other types of traffic. Since inter-datacenter networks are usually overprovisioned, unlike prior work that mainly aims to maximize link utilization, we propose a delay-optimized traffic routing scheme to explicitly differentiate path selection for different sessions according to their delay sensitivities, leading to a software-defined inter-datacenter networking overlay implemented at the application layer. We show that our solution can yield sparse path selection by only solving linear programs, and thus, in contrast to prior traffic engineering solutions, does not lead to overly fine-grained traffic splitting, further reducing packet resequencing overhead and the number of forwarding rules to be installed in each forwarding unit. Real-world experiments based on a deployment on six globally distributed Amazon EC2 datacenters have shown that our system can effectively prioritize and improve the delay performance of inter-datacenter video flows at a low cost.

Control plane innovations to realize dynamic formulation of multicast sessions in inter-DC software-defined elastic optical networks

It is known that to support the applications such as datacenter backup and migration, multicast should be supported efficiently in inter-datacenter (inter-DC) networks to carry the corresponding point-to-multiple-point communications. Moreover, due to the traffic dynamics in inter-DC networks, we might have to consider the case that the multicast members can join or leave a multicast session dynamically. Therefore, in this work, we try to leverage control plane innovations to realize dynamic formulation of multicast sessions in inter-DC software-defined elastic optical networks (SD-EONs), which are equipped with multicast-incapable bandwidth-variable wavelength selective switches (MI-BV-WSS'). Here, one key issue to address is that the continuous changing of multicast group members can degrade the optimality of a multicast-tree. Hence, we propose to rearrange the multicast-trees adaptively to reduce their spectrum usage. Meanwhile, we try to minimize the frequency of rearrangements to avoid unnecessary operation complexity. Based on these considerations, we propose several multicast-tree rearrangement algorithms for updating multicast sessions dynamically with lightpath reroutings in inter-DC SD-EONs. Both partial and full multicast-tree rearrangements are studied. Simulation results indicate that the proposed algorithms can rearrange the multicast-trees intelligently such that the blocking probability can be reduced effectively with the least lightpath reroutings. Next, based on these theoretical investigations, we consider how to implement the proposed algorithms in the control plane of an inter-DC SD-EON. We extend the OpenFlow (OF) protocol to support the dynamic formulation of multicast sessions and also design the functional models in the control plane elements to realize multicast-tree rearrangements. Experiment results verify the effectiveness of our proposed algorithms and system design.

A General Communication Cost Optimization Framework for Big Data Stream Processing in Geo-Distributed Data Centers

With the explosion of big data, processing large numbers of continuous data streams, i.e., big data stream processing (BDSP), has become a crucial requirement for many scientific and industrial applications in recent years. By offering a pool of computation, communication and storage resources, public clouds, like Amazon's EC2, are undoubtedly the most efficient platforms to meet the ever-growing needs of BDSP. Public cloud service providers usually operate a number of geo-distributed datacenters across the globe. Different datacenter pairs are with different inter-datacenter network costs charged by Internet Service Providers (ISPs). While, inter-datacenter traffic in BDSP constitutes a large portion of a cloud provider's traffic demand over the Internet and incurs substantial communication cost, which may even become the dominant operational expenditure factor. As the datacenter resources are provided in a virtualized way, the virtual machines (VMs) for stream processing tasks can be freely deployed onto any datacenters, provided that the Service Level Agreement (SLA, e.g., quality-of-information) is obeyed. This raises the opportunity, but also a challenge, to explore the inter-datacenter network cost diversities to optimize both VM placement and load balancing towards network cost minimization with guaranteed SLA. In this paper, we first propose a general modeling framework that describes all representative inter-task relationship semantics in BDSP. Based on our novel framework, we then formulate the communication cost minimization problem for BDSP into a mixed-integer linear programming (MILP) problem and prove it to be NP-hard. We then propose a computation-efficient solution based on MILP. The high efficiency of our proposal is validated by extensive simulation based studies.

Simplified DSP-Based Signal–Signal Beat Interference Mitigation Technique for Direct Detection OFDM

Short- and medium-haul links in interdata center, access, and metro networks require cost-effective direct-detection wavelength-division multiplexing transceivers offering energy efficiency, high information spectral density (ISD), and dispersion tolerance. Single-sideband orthogonal frequency-division multiplexing (SSB-OFDM) with direct detection is a potential solution; however, it suffers a penalty from signal–signal beat interference (SSBI) caused by the square-law photodetection. In this paper, a novel DSP-based SSBI mitigation technique, with lower complexity than previously proposed methods, is proposed and assessed through numerical simulations for the first time. The performance improvement is quantified by simulations of 9 × 112 Gb/s 16-QAM SSB-OFDM signal with a net optical ISD of 2.1 (b/s)/Hz. The performance is shown to be similar to that of the more complex receiver-based iterative SSBI compensation technique. Simulations predict an 8.7 dB reduction in the required OSNR at the 7% overhead HD-FEC threshold, and increases up to 100% in maximum reach over uncompensated standard single-mode fibre using the proposed simplified SSBI compensation technique.

Availability-Aware Survivable Virtual Network Embedding in Optical Datacenter Networks

In this work, we study the availability-aware survivable virtual network embedding (A-SVNE) problem in optical interdatacenter networks that use wavelength-division multiplexing. With A-SVNE, we try to satisfy the availability requirement of each virtual component (i.e., a virtual link or a virtual node) in a virtual network. We first analyze the availability of a virtual component based on the availabilities of the substrate link(s) and node(s). Then, we formulate an integer linear programming model for the A-SVNE problem and propose several time-efficient heuristics. Specifically, we design two node mapping strategies: one is sequential selection using efficient weights defined by the availability information, while the other uses auxiliary graphs to transform the problem into a classical problem in graph theory, i.e., the maximum-weight maximum clique. Finally, we use extensive simulations to compare the proposed A-SVNE algorithms with existing ones in terms of the blocking probability, availability gap, and penalty due to service-level agreement violations, and the results indicate that our algorithms perform better.

Dynamic Virtual Network Embedding Over Multilayer Optical Networks

Network virtualization is meant to improve the efficiency of network infrastructure by sharing a physical substrate network among multiple virtual networks. Virtual network embedding (VNE) determines how to map a virtual network request onto a physical substrate. In this paper, we first overview three possible underlying substrates for interdatacenter networks, namely an electrical-layer-based substrate, an optical-layer-based substrate, and a multilayer-based (optical and electrical layer) substrate. Then, the corresponding VNE problems for the three physical substrates are discussed. The work presented focuses on VNE over a multilayer optical network; a key problem is how to map a virtual network request onto either an electrical or optical substrate. We propose an auxiliary graph model to address multilayer virtual network mapping in a dynamic traffic scenario. Different node-mapping and link-mapping policies can be achieved by adjusting weights of the edges of the auxiliary graph, which depends on the purposes of the network operators.

Availability-aware service provisioning in SD-EON-based inter-datacenter networks

The tremendous transmission capacity and flexible spectrum allocation scheme make elastic optical networks (EONs) one of the most promising infrastructures for constructing the interconnection in datacenter (DC) networks. Meanwhile, as DC traffics exhibit highly dynamic and heterogeneous features, differentiated service provisioning schemes are desired. In this paper, we take the advantage of centralized network control and management provided by the software-defined elastic optical networks (SD-EONs) and investigate availability-aware service provisioning in SD-EON-based inter-datacenter (inter-DC) networks. We first describe the problem of availability-aware service provisioning in SD-EON-based inter-DC networks and present the theoretical analysis for service availability. Then, we propose an availability-aware service provisioning algorithm (ASP) that leverages different path protection schemes to satisfy different service availability requirements. A service downgrading (SD) strategy is also designed as a supplement of ASP to further improve its performance. Simulation results show that the proposed ASP-SD algorithm can effectively improve the spectrum efficiency without sacrificing availability.

On the future integrated datacenter networks: Designs, operations, and solutions

Inter-datacenter networks (Inter-DCNs) and Intra-datacenter networks (Intra-DCNs) are entirely different in communication scales, architecture topologies and sizes, and characteristics of traffic flows. An Inter-DCN is built upon the existing Internet backbone with a fixed mesh topology and is hence difficult to accommodate any physical update or change without significant human intervention. Several core nodes of the backbone network are connecting to Intra-DCNs forming the so-called datacenter (DC) nodes. Different Inter-DCN or Intra-DCN architectures imply different network constructions overhead. An optimized combination of Inter-DCN and Intra-DCN architecture is expected to minimize the construction complexity and cost of DCNs while maintaining efficient network performance. However, research work in this important area has received little attention. In this paper, we study the construction of integrated Inter- and Intra- datacenter networks (integrated-DCN). Specifically, by taking advantages of different optical switching technologies and embedding the new control technology of software defined networking (SDN) in the control plane of integrated-DCN, novel cost-effective DCN architecture will be developed. For the proposed integrated architecture, the important characteristic of traffic asymmetry is investigated and accommodated by developing asymmetric network configuration based on SDN-assisted traffic monitoring/prediction, and the issue of network survivability/availability is explored and addressed by taking into consideration of the unique feature of DCN traffic.