Large-scale Trace-driven Simulations Research Articles

Bottleneck-Aware Non-Clairvoyant Coflow Scheduling With Fai

Coflow scheduling is critical to data-parallel applications in data centers. While schemes like Varys can achieve optimal performance, they require a priori information about coflows which is hard to obtain in practice. Existing non-clairvoyant solutions like Aalo generalize least attained service (LAS) scheduling discipline to address this issue. However, they fail to identify the bottleneck flows in a coflow and tend to allocate excessive bandwidth to the non-bottleneck flows, leading to bandwidth wastage and inferior overall performance. To this end, we present Fai that strives to improve the overall coflow performance by accelerating the bottleneck flows without priori knowledge. Fai employs bottleneck-aware scheduling. It adopts loose coordination to update coflow priority and flow rates based on total bytes sent. In addition, Fai detects bottleneck flows based on a flow’s rate and bytes sent, and de-allocates bandwidth for other flows to match the bottleneck rate without affecting the coflow completion time (CCT). The saved bandwidth is then distributed among coflows according to their priority to improve overall performance. Testbed evaluation on a 40-node cluster shows that Fai improves average (P95) CCT by 1.73× (3.43×), compared to Aalo. Large-scale trace-driven simulations also show that Fai outperforms Aalo substantially.

Efficient Online Scheduling for Coflow-Aware Machine Learning Clusters

Distributed machine learning (DML) is an increasingly important workload. In a DML job, each communication phase can comprise a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">coflow</i> , and there are dependencies among its coflows. Thus, efficient coflow scheduling becomes critical for DML jobs. However, the majority of existing solutions focus on scheduling single-stage coflows with no dependencies. While there are a few studies schedule dependent coflows of multi-stage jobs, they suffer from either practical or theoretical issues. Motivated by this situation, we study how to schedule dependent coflows of multiple DML jobs to minimize the total JCT in a shared cluster. We present a formal mathematical formulation for this problem and prove its NP-hardness. To solve this problem without job size information, we present an online coflow-aware optimization framework called <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Parrot</i> . The core idea in <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Parrot</i> is to infer the job with the shortest remaining processing time (SRPT) each time and dynamically control the inferred job's bandwidth based on how confident it is an SRPT job while being mindful of not starving any other job. Specifically, in the design of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Parrot</i> , we present a least per-coflow attained service (LPCAS) policy to infer the SRPT job. We further propose a dynamic job weight assignment mechanism and a linear program (LP) based weighted bandwidth scaling strategy for sharing bandwidth among DML jobs. We have proved that <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Parrot</i> algorithm has a non-trivial competitive ratio. The results from large-scale trace-driven simulations further demonstrate that our <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Parrot</i> can reduce the total JCT by up to 58.4 percent, compared to the state-of-the-art Aalo solution.

Low Latency Big Data Processing without Prior Information

Job scheduling plays an important role in improving the overall system performance in big data processing frameworks. Simple job scheduling policies, such as Fair and FIFO scheduling, do not consider job sizes and may degrade the performance when jobs of varying sizes arrive. More elaborate job scheduling policies make the convenient assumption that jobs are recurring, and complete information about their sizes is available from their prior runs. In this paper, we design and implement an efficient and practical job scheduler for big data processing systems to achieve better performance even without prior information about job sizes. The superior performance of our job scheduler originates from the design of multiple level priority queues, where jobs are demoted to lower priority queues if the amount of service consumed so far reaches a certain threshold. In this case, jobs in need of a small amount of service can finish in the topmost several levels of queues, while jobs that need a large amount of service to complete are moved to lower priority queues to avoid head-of-line blocking. Our new job scheduler can effectively mimic the shortest job first scheduling policy without knowing the job sizes in advance. To demonstrate its performance, we have implemented our new job scheduler in YARN, a popular resource manager used by Hadoop/Spark, and validated its performance with experiments in both real testbeds including Amazon EC2 and large-scale trace-driven simulations. Our experimental and simulation results have strongly confirmed the effectiveness of our design: our new job scheduler can reduce the average job response time of the Fair scheduler by up to 45 percent and achieve better fairness at the same time.

When Crowd Meets Big Video Data: Cloud-Edge Collaborative Transcoding for Personal Livecast

Deep penetration of personal computing devices and high-speed Internet has enabled everyone to be a broadcaster. In this crowdsourced live streaming service, numerous amateur broadcasters lively stream their video contents to viewers around the world. Consequently, these broadcasters generate a massive amount of video data. The set of video sources and recipients are big as well, so are demand for the storage and computational resources. Transcoding becomes a must to better service these viewers with different network and device configurations. However, the massive amount of video data contributed by countless channels even makes cloud significantly expensive for providing transcoding services to the whole community. In this paper, inspired by the paradigm of Edge Computing, we propose a Cloud-edge collaborative system which combines the idle end-viewers' resources with the cloud to transcode the massive amount of videos at scale. Specifically, we put forward tailored viewer selection algorithms after empirically analyses the viewer behavior data. In the meantime, we propose auction-based payment schemes to motivate these viewers participating in the transcoding. Large-scale trace-driven simulations demonstrate the superiority of our approach in cost reduction and service stability. We further implement a prototype in PlanetLab to prove the feasibility of our design.

Inter-Client Interference Cancellation for Full-Duplex Networks With Half-Duplex Clients

Recent studies have experimentally shown the gains of full-duplex radios. However, due to its relatively higher cost and complexity, we can envision a more practical step in the network evolution is to have a full-duplex access point (AP) but keep the clients half-duplex. Unfortunately, the full-duplex gains can hardly be extracted in practice as the uplink transmission from a half-duplex client introduces inter-client interference to another downlink client. To address this issue, we present the design and implementation of IC2 (Inter-Client Interference Cancellation), the first physical layer solution that exploits the AP’s full-duplex capability to actively cancel the interference at the downlink client. Such active cancellation not only improves the achievable capacity, but also better tolerates imperfect user pairing, simplifying the MAC design as a result. We build a prototype of IC2 on USRP-N200 and evaluate its performance via both testbed experiments and large-scale trace-driven simulations. The results show that, without IC2 , about 60% of client pairs produce no gain from full-duplex transmissions, while, with IC2 the median gain of the achievable rate over conventional half-duplex networks can be $1.65\times $ and $1.47\times $ for 1- and 2-antenna scenarios, respectively, even when clients are simply paired randomly.

&lt;italic&gt;TrafficShaper:&lt;/italic&gt; Shaping Inter-Datacenter Traffic to Reduce the Transmission Cost

The emerging deployment of geographically distributed data centers (DCs) incurs a significant amount of data transfers over the Internet. Such transfers are typically charged by Internet service providers with the widely adopted $q$ th percentile charging model . In such a charging model, the time slots with top ( $100-q$ ) percent of data transmission do not affect the total transmission cost and can be viewed as “free.” This brings the opportunity to optimize the scheduling of inter-DC transfers to minimize the entire transmission cost. However, a very little work has been done to exploit those “free” time slots for scheduling inter-DC transfers. The crux is that existing work either lacks a mechanism to accumulate traffic to “free” time slots, or inevitably relies on prior knowledge of future traffic arrival patterns. In this paper, we present TrafficShaper , a new scheduler that shapes the inter-DC traffic to exploit the “free” time slots involved in the $q$ th percentile charging model , so as to reduce or even minimize the transmission cost. When shaping traffic, TrafficShaper advocates a simple principle: more traffic peaks should be scheduled in “free” time slots, while less traffic differentiation should be maintained among the remaining time slots. To this end, TrafficShaper designs a pricing-aware control framework, which makes online decisions for inter-DC transfers without requiring a prior knowledge of traffic arrivals. To verify the performance of TrafficShaper , we conduct rigorous theoretical analysis based on Lyapunov optimization techniques, large-scale trace-driven simulations, and small-scale testbed implementation. Results from rigorous mathematical analyses demonstrate that TrafficShaper can make the transmission cost arbitrarily close to the optimum value. Extensive trace-driven simulation results show that TrafficShaper can reduce the transmission cost by up to 40.23%, compared with the state-of-the-art solutions. The testbed experiments further verify that TrafficShaper can realistically reduce the transmission cost by up to 19.38%.

FDoF: Enhancing Channel Utilization for 802.11ac

Multi-user multiple input multiple output (MU-MIMO) enables a multi-antenna access point to serve multiple users simultaneously, and has been adopted as the IEEE 802.11ac standard. While several PHY-MAC designs have recently been proposed to improve the throughput performance of a MU-MIMO WLAN, they, however, usually assume that all the concurrent streams are of roughly equal length. In reality, users usually have frames with heterogeneous lengths even after aggregation, leading to different lengths of a transmission time. Hence, the concurrent transmission opportunities might not always be fully utilized when some streams finish earlier than the others in a transmission opportunity. To resolve this inefficiency, this paper presents full degree-of-freedom (FDoF), a PHY-MAC design that exploits a novel power allocation scheme to reduce the idle channel time and further leverages frame padding to better utilize the spatial multiplexing gain. Unlike traditional MIMO power allocation, which aims at maximizing the theoretical sum-rate, FDoF’s power allocation explicitly considers heterogeneous frame lengths and minimizes the channel time required to finish concurrent frames, as a result improving the effective throughput. FDoF’s padding protocol then identifies proper users to reuse the remaining idle channel time, while preventing this padding from harming all the ongoing streams. Our evaluation via large-scale trace-driven simulations demonstrates that FDoF’s improves the throughput by up to $2.83\times $ , or by $1.36\times $ on average, as compared to the conventional 802.11ac. By combining FDoF’s power allocation with frame padding, the average throughput gain can be further increased to $1.75\times $ .

IASK: A Distributed Q&amp;A System Incorporating Social Community and Global Collective Intelligence

Traditional web-based Question and Answer (Q&A) websites cannot easily solve non-factual questions to match askers’ preference. Recent research efforts begin to study social-based Q&A systems that rely on an asker's social friends to provide answers. However, this method cannot find answerers for a question not belonging to the asker's interests. To solve this problem, we propose a distributed Q&A system incorporating both social community intelligence and global collective intelligence, named as iASK. iASK improves the response latency and answer quality in both the social domain and global domain. It uses a neural network based friend ranking method to identify answerer candidates by considering social closeness and Q&A activities. To efficiently identify answerers in the global user base, iASK builds a virtual server tree that embeds the hierarchical structure of interests, and also maps users to the tree based on user interests. To accurately locate the cooperative experts, iASK has a fine-grained reputation system to evaluate user reputation based on their cooperativeness and expertise, and uses a reputation based reward strategy to encourage users to be cooperative. To further improve the performance of iASK, we propose a weak tie assisted social based potential answerer location algorithm and an interest coefficient based uncategorized question forwarding algorithm. To further improve the response quality and cooperativeness, we propose a reputation based reward strategy that motivates users to answer questions from unknown users. Experimental results from large-scale trace-driven simulation and real-world daily usages of the iASK prototype show the superior performance of iASK. It achieves high answer quality with 24 percent higher accuracy, short response latency with 53 percent less delay and effective cooperative incentives with 16 percent more answers compared to other social-based Q&A systems. The results also show the effectiveness of the enhancement algorithms in improving the performance of iASK.

Path Reconstruction in Dynamic Wireless Sensor Networks Using Compressive Sensing

This paper presents CSPR, a compressive-sensing-based approach for path reconstruction in wireless sensor networks. By viewing the whole network as a path representation space, an arbitrary routing path can be represented by a path vector in the space. As path length is usually much smaller than the network size, such path vectors are sparse, i.e., the majority of elements are zeros. By encoding sparse path representation into packets, the path vector (and thus the represented routing path) can be recovered from a small amount of packets using compressive sensing technique. CSPR formalizes the sparse path representation and enables accurate and efficient per-packet path reconstruction. CSPR is invulnerable to network dynamics and lossy links due to its distinct design. A set of optimization techniques is further proposed to improve the design. We evaluate CSPR in both testbed-based experiments and large-scale trace-driven simulations. Evaluation results show that CSPR achieves high path recovery accuracy (i.e., 100% and 96% in experiments and simulations, respectively) and outperforms the state-of-the-art approaches in various network settings.

Fair Network Bandwidth Allocation in IaaS Datacenters via a Cooperative Game Approach

With wide application of virtualization technology, tenants are able to access isolated cloud services by renting the shared resources in Infrastructure-as-a-Service (IaaS) datacenters. Unlike resources such as CPU and memory, datacenter network, which relies on traditional transport-layer protocols, suffers unfairness due to a lack of virtual machine (VM)-level bandwidth guarantees. In this paper, we model the datacenter bandwidth allocation as a cooperative game, toward VM-based fairness across the datacenter with two main objectives: 1) guarantee bandwidth for VMs based on their base bandwidth requirements, and 2) share residual bandwidth in proportion to the weights of VMs. Through a bargaining game approach, we propose a bandwidth allocation algorithm, Falloc, to achieve the asymmetric Nash bargaining solution (NBS) in datacenter networks, which exactly meets our objectives. The cooperative structure of the algorithm is exploited to develop an online algorithm for practical real-world implementation. We validate Falloc with experiments under diverse scenarios and show that by adapting to different network requirements of VMs, Falloc can achieve fairness among VMs and balance the tradeoff between bandwidth guarantee and proportional bandwidth sharing. Our large-scale trace-driven simulations verify that Falloc achieves high utilization while maintaining fairness among VMs in datacenters.

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.

Underground coal mine monitoring with wireless sensor networks

Environment monitoring in coal mines is an important application of wireless sensor networks (WSNs) that has commercial potential. We discuss the design of a Structure-Aware Self-Adaptive WSN system, SASA. By regulating the mesh sensor network deployment and formulating a collaborative mechanism based on a regular beacon strategy, SASA is able to rapidly detect structure variations caused by underground collapses. We further develop a sound and robust mechanism for efficiently handling queries under instable circumstances. A prototype is deployed with 27 mica2 motes in a real coal mine. We present our implementation experiences as well as the experimental results. To better evaluate the scalability and reliability of SASA, we also conduct a large-scale trace-driven simulation based on real data collected from the experiments.