Quality and cost are two key considerations for video conferencing services. Service providers face a dilemma when selecting network tiers to build their infrastructure---relying on Internet links has poor quality, while using premium links brings excessive cost.

In modern multi-tenant data centers, each tenant desires reassuring dependability from the virtualized network fabric – bandwidth guarantee with work conservation, bounded tail latency and resilient reachability. However, the slow convergence of prior works under network dynamics and uncertainties can hardly provide the dependability for tenants. Further, state-of-the-art load balance schemes are guarantee-agnostic and bring great risks on breaking bandwidth guarantee, which is overlooked in prior works. In this paper, we propose vFab, a dependable virtualized fabric framework which can (1) quickly detect network failure in data plane, (2) explicitly select proper paths for all flows, and (3) converge to ideal bandwidth allocation at sub-millisecond. The core idea of vFab is to leverage the programmable data plane to build a fusion of an active edge (e.g., NIC) and an informative core (e.g., switch), where the core sends link status and tenant information to the edge via telemetry to help the latter make a timely and accurate decision on path selection and traffic admission. We fully implement vFab with commodity SmartNICs and programmable switches. Extensive evaluations show that vFab can keep bandwidth guarantee with high bandwidth utilization, low and bounded latency, and resilient reachability under various network scenarios with limited overhead. Application-level experiments show that vFab can improve QPS by$2.4\times $and cut tail latency by$10\times $compared to the alternatives.

In multi-tenant data centers, each tenant desires reassuring predictability from the virtual network fabric - bandwidth guarantee, work conservation, and bounded tail latency. Achieving these goals simultaneously relies on rapid and precise traffic admission. However, the slow convergence (tens of milliseconds) of prior works can hardly satisfy the increasingly rigorous performance demand under dynamic traffic patterns. Further, state-of-the-art load balance schemes are all guarantee-agnostic and bring great risks on breaking bandwidth guarantee, which is overlooked in prior works.

This paper presents the two generations of storage network stacks that reduced the average I/O latency of Alibaba Cloud's EBS service by 72% in the last five years: Luna, a user-space TCP stack that corresponds the latency of network to the speed of SSD; and Solar, a storage-oriented UDP stack that enables both storage and network hardware accelerations.

Network performance anomalies (NPAs), e.g. long-tailed latency, bandwidth decline, etc., are increasingly crucial to cloud providers as applications are getting more sensitive to performance. The fundamental difficulty to quickly mitigate NPAs lies in the limitations of state-of-the-art network monitoring solutions --- coarse-grained counters, active probing, or packet telemetry either cannot provide enough insights on flows or incur too much overhead. This paper presents NetSeer, a flow event telemetry (FET) monitor which aims to discover and record all performance-critical data plane events, e.g. packet drops, congestion, path change, and packet pause. NetSeer is efficiently realized on the programmable data plane. It has a high coverage on flow events including inter-switch packet drop/corruption which is critical but also challenging to retrieve the original flow information, with novel intra- and inter-switch event detection algorithms running on data plane; NetSeer also achieves high scalability and accuracy with innovative designs of event aggregation, information compression, and message batching that mainly run on data plane, using switch CPU as complement. NetSeer has been implemented on commodity programmable switches and NICs. With real case studies and extensive experiments, we show NetSeer can reduce NPA mitigation time by 61%-99% with only 0.01% overhead of monitoring traffic.

In memory computing and high-end distributed storage demand low latency, high throughput, and zero data loss simultaneously from datacenter networks. Existing reactive congestion control approaches cannot both minimize queuing latency and ensure zero data loss. A token-oriented proactive approach can achieve them together by controlling congestion even before sending data packets. However, state-of-the-art token-oriented approaches only strive to optimize network-level metrics: maximizing throughput while achieving flow-level fairness. This article answers the question of how to support objective-aware traffic scheduling in token-oriented approaches. The novelty of Token-Oriented in-network Prioritization (TOP) is that it prioritizes tokens instead of data packets. We make three contributions. Via simulations over a hypothetical TOP system, our first contribution is demonstrating the potential performance gain that can be brought by TOP. Second, we investigate the applicability of TOP. Although the overhead of enabling necessary TOP features in switches is trivial, we find that mainstream commodity datacenter switches do not support them. We hence propose a readily-deployable remedy to achieve in-network prioritization by pushing both switch and end-host hardware capacity to an extreme end. Lastly, we implement a running TOP system with Linux hosts and commodity switches, and evaluate TOP in testbeds and with large-scale simulations for various scenarios.

Remote Direct Memory Access(RDMA) technology rapidly changes the landscape of nowadays datacenter applications. Congestion control for RDMA networking is a critical challenge. As an end-to-end layer 3 congestion control mechanism, Datacenter QCN (DCQCN) alleviates the unfairness and head-of-the-line blocking problems of Priority-based Flow Control (PFC). However, a lossless network does not guarantee low latency even with DCQCN enabled. When network congestion happens, switch queues still build-up due to the response latency of end-to-end solutions. In this article, we propose Predictive PFC (P-PFC) to reduce tail latency in RDMA networks. P-PFC monitors the derivative of buffer occupation, predicts the happening of PFC trigger in the future, and proactively triggers PFC pause in advance. The benefit is that buffer usage can be maintained at a low level, hence the tail latency can be controlled. Preliminary evaluation results demonstrate that P-PFC can reduce tail latency by more than half of that in standard PFC in many scenarios, without hurting the throughput and average latency. P-PFC can also protect innocent flows compared with standard PFC according to our experiments. To our best knowledge, this is the first work of using derivative to improve PFC in lossless RDMA networks.

Image saliency detection has recently witnessed rapid progress due to deep neural networks. However, there still exist many important problems in the existing deep learning based methods. Pixel-wise convolutional neural network (CNN) methods suffer from blurry boundaries due to the convolutional and pooling operations. While region-based deep learning methods lack spatial consistency since they deal with each region independently. In this paper, we propose a novel salient object detection framework using a superpixelwise variational autoencoder (SuperVAE) network. We first use VAE to model the image background and then separate salient objects from the background through the reconstruction residuals. To better capture semantic and spatial contexts information, we also propose a perceptual loss to take advantage from deep pre-trained CNNs to train our SuperVAE network. Without the supervision of mask-level annotated data, our method generates high quality saliency results which can better preserve object boundaries and maintain the spatial consistency. Extensive experiments on five wildly-used benchmark datasets show that the proposed method achieves superior or competitive performance compared to other algorithms including the very recent state-of-the-art supervised methods.

As a hot topic in computer vision, recent researches on salient object detection (SOD) have focused on using the over-designed deep convolutional neural networks (CNNs) to improve the detection accuracy. However, these complex architectures constraint themselves to low speed and drag them on wide-ranging applications. In this paper, we simplify the over-designed networks and propose the Two-Branch Network for Real-time Salient Object Detection (Two-B-Real Net). Particularly, the Perceptual Branch and the Objectness Branch in our network can efficiently capture detailed information and distinctive objectness simultaneously. And we also design novel attention mechanisms to guide the network to focus on most saliency-related features and generate more accurate results. Extensive evaluations show that the proposed algorithm achieves the leading accuracy performance with real-time speed (125fps) which is significantly faster than the existing methods.

Point cloud segmentation is a key problem of 3D multimedia signal processing. Existing methods usually use a single network structure which is trained by a per-point loss. These methods mainly focus on the geometric similarity between the prediction results and the ground truth, ignoring visual perception difference. In this paper, we present a segmentation adversarial network to overcome the drawbacks above. A discriminator is introduced to provide a perceptual loss to increase the rationality judgment of prediction and guide the further optimization of the segmentator. In order to perfectly capture the structural information of parts in the same category of objects, condition settings are employed to add a global constraint. Experimental results show the proposed methods can correct the common errors in point cloud segmentation and obtain more accurate and better segmentation of visual perceptual.

Effective feature representations which should not only express the images individual properties, but also reflect the interaction among group images are essentially crucial for robust co-saliency detection. This paper proposes a novel deep learning co-saliency detection approach which simultaneously learns single image properties and robust group feature in a recurrent manner. Specifically, our network first extracts the semantic features of each image. Then, a specially designed Recurrent Co-Attention Unit (RCAU) will explore all images in the group recurrently to generate the final group representation using the co-attention between images, and meanwhile suppresses noisy information. The group feature which contains complementary synergetic information is later merged with the single image features which express the unique properties to infer robust co-saliency. We also propose a novel co-perceptual loss to make full use of interactive relationships of whole images in the training group as the supervision in our end-to-end training process. Extensive experimental results demonstrate the superiority of our approach in comparison with the state-of-the-art methods.

Software-Defined Networks (SDNs) introduce great flexibilities in how packet routes can be defined and changed over time, and enable a more fine-grained and adaptive traffic engineering. The recently introduced support for more accurate synchronization in SDNs further improves the degree of control an operator can have over the packets' forwarding paths, and also allows to avoid disruptions and inconsistencies during network updates, i.e., during the rerouting of flows. However, how to optimally exploit such technology algorithmically - to efficiently schedule the update of multiple flows in such timed SDNs - while accounting for possible interference and congestion, is not well-understood today. We, in this paper, initiate the study of the fundamental problem of how to reroute the updates of multiple network flows in a synchronized SDN in a congestion-free manner. We rigorously prove that the problem is NP-hard for flows of unit size and network links with unit delay. We also show that a greedy approach to update the network can delay the update significantly. Our main contribution is the first solution to this problem: Chronicle. Our approach is based on time-extended network construction and the resource dependency graph, which is implemented by Openflow 1.5 using the scheduled bundles feature. The evaluation results show that Chronicle can reduce the makespan by 63% and reduce the number of changed rules by 50% compared to state-of-the-art.

The advent of more accurate synchronization in Software-Defined Networks (SDNs) in general and the notion of timed updates in particular, enables operators to fully exploit the potential of the more fine-grained and adaptive traffic engineering, by avoiding disruptions and inconsistencies during the update. However, little is known today about how to schedule the update of multiple flows in such timed SDNs: As flows compete for limited resources, implementing a congestion-free update remains algorithmically challenging, even in timed SDNs. This paper initiates the study of the fundamental problem of how to reroute the update of multiple network flows in a synchronized SDN in a congestion-free manner. We show that that the problem is NP-hard already for flows of unit size and network links with unit delay. Our main contribution is a first solution for this problem: Chronicle. Our approach is based on a time-extended network construction and resource dependency graph, which is implemented by Openflow 1.5 using the scheduled bundles feature. Evaluation results show that Chronicle can reduce the makespan by 63% and reduce the number of changed rules by 50% compared to state-of-the-art.

State-of-the-art inter-datacenter WANs rely on software defined networking (SDN) to orchestrate their data transmission. Optimization requires frequent network update operations to switch forwarding tables. When scheduling inter-datacenter WANs, the utility of services should be respected. Yet, existing network update approaches do not respect network utility and could result in performance degradation during the network update procedure. Further, the update causes not only performance degradation, but also the degradation period is unnecessarily prolonged. In this paper we propose Hermes, a utility-aware network update system. We aim to find a rate limiting scheme for update which maximizes the sum of service utility, while ensuring the congestion-free property during the update. We propose an optimization framework for the maximum utility network update problem (MUP). MUP is NP-hard and a series of algorithms are developed to solve it. Extensive simulation and testbed experiments with a prototype demonstrate that Hermes can increase the total utility by 80% compared to state-of-the-art. At the same time, it reduces the total update time and control overhead by 40% and 55%, respectively.