arXiv subjects

Hong Xu

Publications and source records attributed to Hong Xu.

At least 91 records · Page 5Linked to original sources

arXiv2019

Scalable Traffic Engineering for Higher Throughput in Heavily-loaded Software Defined Networks

Existing traffic engineering (TE) solutions performs well for software defined network (SDN) in average cases. However, during peak hours, bursty traffic spikes are challenging to handle, because it is difficult to react in time and guarantee high performance even after failures with limited flow entries. Instead of leaving some capacity empty to guarantee no congestion happens due to traffic rerouting after failures or path updating after demand or topology changes, we decide to make full use of the network capacity to satisfy the demands for heavily-loaded peak hours. The TE system also needs to react to failures quickly and utilize the priority queue to guarantee the transmission of loss and delay sensitive traffic. We propose TED, a scalable TE system that can guarantee high throughput in peak hours. TED can quickly compute a group of maximum number of edge-disjoint paths for each ingress-egress switch pair. We design two methods to select paths under the flow entry limit. We then input the selected paths to our TE to minimize the maximum link utilization. In case of large traffic matrix making the maximum link utilization larger than 1, we input the utilization and the traffic matrix to the optimization of maximizing overall throughput under a new constrain. Thus we obtain a realistic traffic matrix, which has the maximum overall throughput and guarantees no traffic starvation for each switch pair. Experiments show that TED has much better performance for heavily-loaded SDN and has 10% higher probability to satisfy all (> 99.99%) the traffic after a single link failure for G-Scale topology than Smore under the same flow entry limit.

cs.NI↗

arXiv2019

Model Asset eXchange: Path to Ubiquitous Deep Learning Deployment

A recent trend observed in traditionally challenging fields such as computer vision and natural language processing has been the significant performance gains shown by deep learning (DL). In many different research fields, DL models have been evolving rapidly and become ubiquitous. Despite researchers' excitement, unfortunately, most software developers are not DL experts and oftentimes have a difficult time following the booming DL research outputs. As a result, it usually takes a significant amount of time for the latest superior DL models to prevail in industry. This issue is further exacerbated by the common use of sundry incompatible DL programming frameworks, such as Tensorflow, PyTorch, Theano, etc. To address this issue, we propose a system, called Model Asset Exchange (MAX), that avails developers of easy access to state-of-the-art DL models. Regardless of the underlying DL programming frameworks, it provides an open source Python library (called the MAX framework) that wraps DL models and unifies programming interfaces with our standardized RESTful APIs. These RESTful APIs enable developers to exploit the wrapped DL models for inference tasks without the need to fully understand different DL programming frameworks. Using MAX, we have wrapped and open-sourced more than 30 state-of-the-art DL models from various research fields, including computer vision, natural language processing and signal processing, etc. In the end, we selectively demonstrate two web applications that are built on top of MAX, as well as the process of adding a DL model to MAX.

cs.LG↗

arXiv2019

Flash: Efficient Dynamic Routing for Offchain Networks

Offchain networks emerge as a promising solution to address the scalability challenge of blockchain. Participants directly make payments through a network of payment channels without the overhead of committing onchain transactions. Routing is critical to the performance of offchain networks. Existing solutions use either static routing with poor performance or dynamic routing with high overhead to obtain the dynamic channel balance information. In this paper, we propose Flash, a new dynamic routing solution that leverages the unique characteristics of transactions in offchain networks to strike a better tradeoff between path optimality and probing overhead. By studying the traces of real offchain networks, we find that the payment sizes are heavy-tailed, and most payments are highly recurrent. Flash thus differentiates the treatment of elephant payments from mice payments. It uses a modified max-flow algorithm for elephant payments to find paths with sufficient capacity, and strategically routes the payment across paths to minimize the transaction fees. Mice payments are directly sent by looking up a routing table with a few precomputed paths to reduce probing overhead. Testbed experiments and data-driven simulations show that Flash improves the success volume of payments by up to 2.3x compared to the state-of-the-art routing algorithm.

cs.NI↗

arXiv2019

Centrality-based Middlepoint Selection for Traffic Engineering with Segment Routing

Segment routing is an emerging technology to simplify traffic engineering implementation in WANs. It expresses an end-to-end logical path as a sequence of segments through a set of middlepoints. Traffic along each segment is routed along shortest paths. In this paper, we study practical traffic engineering (TE) with segment routing in SDN based WANs. We consider two common types of TE, and show that the TE problem can be solved in weakly polynomial time when the number of middlepoints is fixed and not part of the input. However, the corresponding linear program is of large scale and computationally expensive. For this reason, existing methods that work by taking each node as a candidate middlepoint are inefficient. Motivated by this, we propose to select just a few important nodes as middlepoints for all traffic. We use node centrality concepts from graph theory, notably group shortest path centrality, for middlepoint selection. Our performance evaluation using realistic topologies and traffic traces shows that a small percentage of the most central nodes can achieve good results with orders of magnitude lower runtime.

cs.NI↗

arXiv2019

NetKernel: Making Network Stack Part of the Virtualized Infrastructure

This paper presents a system called NetKernel that decouples the network stack from the guest virtual machine and offers it as an independent module. NetKernel represents a new paradigm where network stack can be managed as part of the virtualized infrastructure. It provides important efficiency benefits: By gaining control and visibility of the network stack, operator can perform network management more directly and flexibly, such as multiplexing VMs running different applications to the same network stack module to save CPU cores, and enforcing fair bandwidth sharing with distributed congestion control. Users also benefit from the simplified stack deployment and better performance. For example mTCP can be deployed without API change to support nginx natively, and shared memory networking can be readily enabled to improve performance of colocated VMs. Testbed evaluation using 100G NICs shows that NetKernel preserves the performance and scalability of both kernel and userspace network stacks, and provides the same isolation as the current architecture.

cs.NI↗

arXiv2019

Saec: Similarity-Aware Embedding Compression in Recommendation Systems

Production recommendation systems rely on embedding methods to represent various features. An impeding challenge in practice is that the large embedding matrix incurs substantial memory footprint in serving as the number of features grows over time. We propose a similarity-aware embedding matrix compression method called Saec to address this challenge. Saec clusters similar features within a field to reduce the embedding matrix size. Saec also adopts a fast clustering optimization based on feature frequency to drastically improve clustering time. We implement and evaluate Saec on Numerous, the production distributed machine learning system in Tencent, with 10-day worth of feature data from QQ mobile browser. Testbed experiments show that Saec reduces the number of embedding vectors by two orders of magnitude, compresses the embedding size by ~27x, and delivers the same AUC and log loss performance.

cs.IR↗

arXiv2019

Stanza: Layer Separation for Distributed Training in Deep Learning

The parameter server architecture is prevalently used for distributed deep learning. Each worker machine in a parameter server system trains the complete model, which leads to a hefty amount of network data transfer between workers and servers. We empirically observe that the data transfer has a non-negligible impact on training time. To tackle the problem, we design a new distributed training system called Stanza. Stanza exploits the fact that in many models such as convolution neural networks, most data exchange is attributed to the fully connected layers, while most computation is carried out in convolutional layers. Thus, we propose layer separation in distributed training: the majority of the nodes just train the convolutional layers, and the rest train the fully connected layers only. Gradients and parameters of the fully connected layers no longer need to be exchanged across the cluster, thereby substantially reducing the data transfer volume. We implement Stanza on PyTorch and evaluate its performance on Azure and EC2. Results show that Stanza accelerates training significantly over current parameter server systems: on EC2 instances with Tesla V100 GPU and 10Gb bandwidth for example, Stanza is 1.34x--13.9x faster for common deep learning models.

cs.LG↗

arXiv2018

Overview: A Hierarchical Framework for Plan Generation and Execution in Multi-Robot Systems

The authors present an overview of a hierarchical framework for coordinating task- and motion-level operations in multirobot systems. Their framework is based on the idea of using simple temporal networks to simultaneously reason about precedence/causal constraints required for task-level coordination and simple temporal constraints required to take some kinematic constraints of robots into account. In the plan-generation phase, the framework provides a computationally scalable method for generating plans that achieve high-level tasks for groups of robots and take some of their kinematic constraints into account. In the plan-execution phase, the framework provides a method for absorbing an imperfect plan execution to avoid time-consuming re-planning in many cases. The authors use the multirobot path-planning problem as a case study to present the key ideas behind their framework for the long-term autonomy of multirobot systems.

cs.AI↗

arXiv2017

Measuring Territorial Control in Civil Wars Using Hidden Markov Models: A Data Informatics-Based Approach

Territorial control is a key aspect shaping the dynamics of civil war. Despite its importance, we lack data on territorial control that are fine-grained enough to account for subnational spatio-temporal variation and that cover a large set of conflicts. To resolve this issue, we propose a theoretical model of the relationship between territorial control and tactical choice in civil war and outline how Hidden Markov Models (HMMs) are suitable to capture theoretical intuitions and estimate levels of territorial control. We discuss challenges of using HMMs in this application and mitigation strategies for future work.

stat.AP↗

arXiv2017

Dynamic Switch-Controller Association and Control Devolution for SDN Systems

In software-defined networking (SDN), as data plane scale expands, scalability and reliability of the control plane have become major concerns. To mitigate such concerns, two kinds of solutions have been proposed separately. One is multi- controller architecture, i.e., a logically centralized control plane with physically distributed controllers. The other is control devolution, i.e., delegating control of some flows back to switches. Most of existing solutions adopt either static switch-controller association or static devolution, which may not adapt well to the traffic variation, leading to high communication costs between switches and controller, and high computation costs of switches. In this paper, we propose a novel scheme to jointly consider both solutions, i.e., we dynamically associate switches with controllers and dynamically devolve control of flows to switches. Our scheme is an efficient online algorithm that does not need the statistics of traffic flows. By adjusting a parameter, we can make a trade- off between costs and queue backlogs. Theoretical analysis and extensive simulations show that our scheme yields much lower costs or latency compared to other schemes, as well as balanced loads among controllers.

cs.NI↗

arXiv2017

Overview: Generalizations of Multi-Agent Path Finding to Real-World Scenarios

Multi-agent path finding (MAPF) is well-studied in artificial intelligence, robotics, theoretical computer science and operations research. We discuss issues that arise when generalizing MAPF methods to real-world scenarios and four research directions that address them. We emphasize the importance of addressing these issues as opposed to developing faster methods for the standard formulation of the MAPF problem.

cs.AI↗

arXiv2016

Improved Algorithms and Coupled Neutron-Photon Transport for Auto-Importance Sampling Method

The Auto-Importance Sampling (AIS) method is a Monte Carlo variance reduction technique proposed for deep penetration problems, which can significantly improve computational efficiency without pre-calculations for importance distribution. However, the AIS method is only validated with several simple examples, and cannot be used for coupled neutron-photon transport. This paper presents the improved algorithms for the AIS method, including particle transport, fictitious particles creation and adjustment, fictitious surface geometry, random number allocation and calculation of the estimated relative error. These improvements allow the AIS method to be applicable to complicated deep penetration problems with complex geometry and multiple materials. A coupled Neutron-Photon Auto-Importance Sampling (NP-AIS) method is proposed to solve the deep penetration problems of coupled neutron-photon transport using the improved algorithms. The NUREG/CR-6115 PWR benchmark was calculated by using the methods of NP-AIS, geometry splitting with Russian roulette and the analog Monte Carlo, respectively. The calculation results of NP-AIS were in good agreement with those of geometry splitting with Russian roulette and the benchmark solutions. The computational efficiency of NP-AIS for both neutron and photon was much better than that of geometry splitting with Russian roulette in most cases, and increased by several orders of magnitude compared with that of the analog Monte Carlo.

nucl-th↗

arXiv2016

Benchmarking NFV Software Dataplanes

A key enabling technology of NFV is software dataplane, which has attracted much attention in both academia and industry recently. Yet, till now there is little understanding about its performance in practice. In this paper, we make a benchmark measurement study of NFV software dataplanes in terms of packet processing capability, one of the most fundamental and critical performance metrics. Specifically, we compare two state-of-the-art open-source NFV dataplanes, SoftNIC and ClickOS, using commodity 10GbE NICs under various typical workloads. Our key observations are that (1) both dataplanes have performance issues processing small (<=128B) packets; (2) it is not always best to put all VMs of a service chain on one server due to NUMA effect. We propose resource allocation strategies to remedy the problems, including carefully adding CPU cores and vNICs to VMs, and spreading VMs of a service chain to separate servers. To fundamentally address these problems and scale their performance, SoftNIC and ClickOS could improve the support for NIC queues and multiple cores.

cs.NI↗

arXiv2016

An Alternating Direction Method Approach to Cloud Traffic Management

In this paper, we introduce a unified framework for studying various cloud traffic management problems, ranging from geographical load balancing to backbone traffic engineering. We first abstract these real-world problems as a multi-facility resource allocation problem, and then present two distributed optimization algorithms by exploiting the special structure of the problem. Our algorithms are inspired by Alternating Direction Method of Multipliers (ADMM), enjoying a number of unique features. Compared to dual decomposition, they converge with non-strictly convex objective functions; compared to other ADMM-type algorithms, they not only achieve faster convergence under weaker assumptions, but also have lower computational complexity and lower message-passing overhead. The simulation results not only confirm these desirable features of our algorithms, but also highlight several additional advantages, such as scalability and fault-tolerance.

cs.NI↗

arXiv2015

Novel Layered Iridate Ba7Ir3O13+δ Thin Films with Colossal Permittivity

Ba7Ir3O13+δ in thin film form is discovered. These films are characterized by colossal permittivity (CP) ~104 at room temperature, attributable to the colossal internal barrier layer capacitance effect at atomically thin domain boundaries. These findings suggest a new route to seeking novel CP materials through designing atomically thin domain boundaries.

cond-mat.mtrl-sci↗

arXiv2015

RepNet: Cutting Tail Latency in Data Center Networks with Flow Replication

Data center networks need to provide low latency, especially at the tail, as demanded by many interactive applications. To improve tail latency, existing approaches require modifications to switch hardware and/or end-host operating systems, making them difficult to be deployed. We present the design, implementation, and evaluation of RepNet, an application layer transport that can be deployed today. RepNet exploits the fact that only a few paths among many are congested at any moment in the network, and applies simple flow replication to mice flows to opportunistically use the less congested path. RepNet has two designs for flow replication: (1) RepSYN, which only replicates SYN packets and uses the first connection that finishes TCP handshaking for data transmission, and (2) RepFlow which replicates the entire mice flow. We implement RepNet on {\tt node.js}, one of the most commonly used platforms for networked interactive applications. {\tt node}'s single threaded event-loop and non-blocking I/O make flow replication highly efficient. Performance evaluation on a real network testbed and in Mininet reveals that RepNet is able to reduce the tail latency of mice flows, as well as application completion times, by more than 50\%.

cs.NI↗

arXiv2014

Low Latency Datacenter Networking: A Short Survey

Datacenters are the cornerstone of the big data infrastructure supporting numerous online services. The demand for interactivity, which significantly impacts user experience and provider revenue, is translated into stringent timing requirements for flows in datacenter networks. Thus low latency networking is becoming a major concern of both industry and academia. We provide a short survey of recent progress made by the networking community for low latency datacenter networks. We propose a taxonomy to categorize existing work based on four main techniques, reducing queue length, accelerating retransmissions, prioritizing mice flows, and exploiting multi-path. Then we review select papers, highlight the principal ideas, and discuss their pros and cons. We also present our perspectives of the research challenges and opportunities, hoping to aspire more future work in this space.

cs.NI↗

arXiv2013

Reducing Electricity Demand Charge for Data Centers with Partial Execution

Data centers consume a large amount of energy and incur substantial electricity cost. In this paper, we study the familiar problem of reducing data center energy cost with two new perspectives. First, we find, through an empirical study of contracts from electric utilities powering Google data centers, that demand charge per kW for the maximum power used is a major component of the total cost. Second, many services such as Web search tolerate partial execution of the requests because the response quality is a concave function of processing time. Data from Microsoft Bing search engine confirms this observation. We propose a simple idea of using partial execution to reduce the peak power demand and energy cost of data centers. We systematically study the problem of scheduling partial execution with stringent SLAs on response quality. For a single data center, we derive an optimal algorithm to solve the workload scheduling problem. In the case of multiple geo-distributed data centers, the demand of each data center is controlled by the request routing algorithm, which makes the problem much more involved. We decouple the two aspects, and develop a distributed optimization algorithm to solve the large-scale request routing problem. Trace-driven simulations show that partial execution reduces cost by $3\%--10.5\%$ for one data center, and by $15.5\%$ for geo-distributed data centers together with request routing.

cs.NI↗