A lot of researches exploit the flexibility of Software-Defined Networking (SDN) to conduct traffic engineering in order to improve network performance and enhance robustness to failures. As the upgrade of a traditional network to a full SDN deployment is an incremental process, the coexistence of SDN switches and legacy switches forms a hybrid SDN. Due to the different forwarding characteristics of these switches, it is essential to coordinate the forwarding of SDN control and distributed routing to avoid inconsistency and achieve high network utilization. In this paper, we note that the effectiveness of traffic engineering in hybrid SDN strongly depends on both the structures of forwarding graphs and traffic distribution, while existing approaches mainly focus on the latter. We first define the consistent forwarding graph, and then construct forwarding graphs with potential high throughput for effective traffic engineering, while maintaining forwarding consistency. The evaluation results show that the proposed forwarding graph construction approach improves network throughput and achieves better load balancing compared with existing simple forwarding path constructing approaches, especially with more fraction of SDN deployment.

The upgrade of a legacy network to a full software-defined networking (SDN) deployment is usually an incremental process, during which SDN switches and legacy switches coexist in the hybrid network. However, with inappropriate deployment of SDN switches and design of hybrid control, the advantages of SDN control could not exert, and it even results in performance degradation or inconsistency (e.g., loops, black-holes). Therefore, the hybrid SDN requires considerable coordination of the centralized control and distributed routing. In this paper, we propose a solution to handle the heterogeneity caused by distinct forwarding characteristics of SDN and legacy switches, therefore boosting the benefits of hybrid SDN. We plan SDN placement to enhance the SDN controllability over the hybrid network, and conduct traffic engineering considering both the forwarding characteristics of SDN and legacy switches. The experiments with various topologies show that the SDN placement planning and hybrid forwarding yield better network performance especially in the early 70% SDN deployment.

With SDN control programs from multi-domains configuring the network, it is inevitable that control programs make conflicting control decisions, which probably lead to misconfiguration or performance degradation. Existing control coordination approaches either compose control programs to derive consistent solutions jointly or examine the generated rules of each control program to ensure they are consistent. However, the former is usually of great complexity and hard to be conducted automatically, and the latter probably results in suboptimal solutions due to the independent execution of control programs. Moreover, these approaches all fail to consider the control utility of control programs. In this paper, we propose Redactor to optimize the consistency and utility of network control in an automatic and dynamic manner. To make network control consistent, we implement SDN control programs with declarative language Prolog, and compose control programs automatically to execute together to make consistent decisions. When conflicts occur, we use a heuristic approach to compromise a subset of control programs to maximize the control utility. We compare Redactor with the static priority mechanism and Athens [1], and the results show that Redactor always satisfies more control objectives to achieve better control consistency and utility.

At the present time, billions of videos are hosted and shared in the cloud of which a sizable portion consists of near-duplicate video copies. An efficient and accurate content-based online near-duplicate video detection method is a fundamental research goal; as it would benefit applications such as duplication-aware storage, pirate video detection, polluted video tag detection, searching result diversification. Despite the recent progress made in near-duplicate video detection, it remains challenging to develop a practical detection system for large-scale applications that has good efficiency and accuracy performance. In this paper, we shift the focus from feature representation design to system design, and develop a novel system, called CompoundEyes, accordingly. The improvement in accuracy is achieved via well-organized classifiers instead of advanced feature design. Meanwhile, by applying simple features with reduced dimensionality and exploiting the parallelism of the detection architecture, we accelerate the detection speed. Through extensive experiments we demonstrate that the proposed detection system is accurate and fast. It takes approximately 1.45 seconds to process a video clip from a large video dataset, CC_WEB_VIDEO, with a 89% detection accuracy.

With the popular applications of SDN in load balancing and failure recovery, the controller schedules affected flows to redundant paths to avoid network congestions and failures by updating flow tables in data plane. However, inconsistent flow table updating may lead to transient incorrect network behaviors or undesired performance degradation. Therefore, the consistency imposes dependencies among updates, so that the order of updates must be carefully considered to keep the consistency. To update flow tables consistently and efficiently, in this paper, we propose an update ordering approach — Cupid. To avoid high overhead in update ordering, we divide the global dependencies among updates into local restrictions by: 1) partitioning a new routing path into several independent segments, 2) identifying critical nodes controlling traffic shifting between the old path and new path, and 3) constructing a dependency graph among critical nodes for potential congested links. We then design a heuristic algorithm to resolve the dependency graph. To save the flow table space, a switch keeps only one flow entry with multiple ports for a flow during updating. Our simulation shows that Cupid schedules updates at least 2 times faster and has less throughput losses than the state-of-the-art approaches in both fat-tree and mesh networks.

Software-defined networking enables applications act as blackboxes independently to control the network flexibly. However, these independent applications may generate conflicting control decisions. To reconcile applications automatically and dynamically, we implement control applications with Prolog, which enables applications to execute jointly to make consistent control decisions. When conflicts occur, we design a compromise algorithm by sacrificing a subset of applications to maximize the desired control objectives.

In traditional networks, it is difficult to manage the distributed detection and prevention nodes of IDS and IPS due to the laborious manual deployment and independent configuration. Software defined networking (SDN) provides a flexible approach to control the underlying network infrastructures efficiently. However, the OpenFlow flow table is too simple to provide complex functions with the match-action style processing. To support more functionalities, in this paper, we propose a middlebox management architecture with SDN - OpenMiddlebox, by extending OpenFlow to support middleboxes with ClickOS virtual machines (VM), so that programmable middleboxes could be deployed and managed in switches with fast booted ClickOS VMs flexibly. We then design automatic deployment and update schemes of network intrusion detection and prevention middleboxes with the centralized controller. The evaluation results show that OpenMiddlebox could manage the distributed middleboxes efficiently and is scalable to large networks, and the centralized control also improves the network intrusion detection and prevention accuracy.

The increasing virtualization in data centers brings growing inter-node communication by running various applications on virtual machines located physically separated. Meanwhile, the virtual machines on a physical server also compete for limited Ethernet interface I/O resources. Load balancing with multipath and multihoming is usually the key to address the bandwidth and Ethernet I/O bottlenecks. Even though various variants of equal cost multipath (ECMP) schemes have been widely applied for load balancing, the equal and fairness assumption of ECMP results in imbalance and underutilization in asymmetric networks without considering network topology and traffic situation. The multihoming solutions usually require protocol modification and peer's support. In this paper, we propose a utilization & topology-aware multipath routing and multihoming scheduling with Software Defined Networking (SDN) in data centers to address these bottlenecks. The utilization & topology-aware multipath routing takes global network situation to avoid congestions and balances utilization of multiple paths. At a multi-homed server end, the multihoming scheduler balances the traffic among multiple Ethernet interfaces and ensures QoS guarantees when accessing the network without changing network stack. We compared our approach with traditional single path and equal cost multipath schemes, and the results showed that the utilization & topology-aware multipath routing and multihoming scheduling achieved much higher network utilization and better load balancing, especially for asymmetric networks.