The advent of IoT and edge computing will lead to massive amounts of data that need to be collected and transmitted to online storage systems. To address this problem, we push data deduplication to the network edge. Specifically, we propose a new technique for collaborative edge-facilitated deduplication (EF-dedup), wherein we partition the resource-constrained edge nodes into disjoint clusters, maintain a deduplication index structure for each cluster using a distributed key-value store and perform decentralized deduplication within those clusters. This is a challenging partitioning problem that addresses a novel tradeoff: edge nodes with highly correlated data may not always be within the same edge cloud, with non-trivial network cost among them. We address this challenge by first formulating an optimization problem to partition the edge nodes, considering both the data similarities across the nodes and the inter-node network cost. We prove that the problem is NP-Hard, provide bounded heuristics to solve it and build a prototype EF-dedup system. Our experiments on EF-dedup, performed on edge nodes in AT&T research lab and a central cloud at AWS, demonstrate that EF-dedup achieves 38.3-118.5% better deduplication throughput than sole cloud-based techniques and achieves 43.4-60.2% lesser aggregate cost in terms of the network-storage tradeoff as compared to approaches that solely favor one over the other.

Data deduplication has become a crucial technique for reducing data in modern storage systems. We present SEP-D, a practical scale-out distributed storage system to incorporate data deduplication for primary storage. SEP-D introduces a novel metadata handling mechanism which combines content-based hashing with built-in distributed data placement strategies such as CRUSH. This enables SEP-D to eliminate the needs for remote metadata lookups, thus incorporating deduplication without affecting scalability. SEP-D integrates smoothly with the existing storage system, allowing the re-use of storage policies across different pools of storage. We implemented SEP-D in Ceph, a popular distributed storage system widely adopted in the industry, and demonstrated that SEP-D has minimal impact on I/O performance in data while maintaining existing storage policies implemented in underlying distributed storage systems.

Erasure-coded storage systems have gained considerable adoption recently since they can provide the same level of reliability with significantly lower storage overhead compared to replicated systems. However, background traffic of such systems – e.g., repair, rebalance, backup and recovery traffic – often has large volume and consumes significant network resources. Independently scheduling such tasks and selecting their sources can easily create interference among data flows, causing severe deadline violation. We show that the well-known heuristic scheduling algorithms fail to consider important constraints, thus resulting in unsatisfactory performance. In this paper, we claim that an optimal scheduling algorithm, which aims to maximize the number of background tasks completed before deadlines, must simultaneously consider task deadline, network topology, chunk placement, and time-varying resource availability. We first show that the corresponding optimization problem is NP-hard. Then we propose a novel algorithm, called Linear Programming for Selected Tasks (LPST) to maximize the number of successful tasks and improve overall utilization of the datacenter network. It jointly schedules tasks and selects their sources based on a notion of Remaining Time Flexibility, which measures the slackness of the starting time of a task. We evaluated the efficacy of our algorithm using extensive simulations and validate the results with experiments in a real cloud environment. Our results show that, under certain scenarios, LPST can perform 7x$\sim$10x better than the heuristics which blindly treat the infrastructure as a collection of homogeneous resources, and 21.7$\sim$65.9 percent better than the algorithms that only take the network topology into account.

Erasure-coded storage systems have gained considerable adoption recently since they can provide the same level of reliability with significantly lower storage overhead compared to replicated systems. However, background traffic of such systems - e.g. repair, rebalance, backup and recovery traffic - often has large volume and consumes significant network resources. Independently scheduling such tasks and selecting their sources can easily create interference among data flows, causing severe deadline violation. We show that the well-known heuristic scheduling algorithms fail to consider important constraints, thus resulting in unsatisfactory performance. In this paper, we claim that an optimal scheduling algorithm that aims to maximizethe number of background tasks completed before deadlines must simultaneously consider deadline-aware scheduling, network topology, chunk placement, and time-varying resource availability. To solve this problem, we propose a novel algorithm, called Linear Programming for Selected Tasks (LPST) to maximize the number of successful tasks and improve overallutilization of the datacenter network. It jointly schedules tasks and selects their sources based on a notion of Remaining Time Flexibility, which measures the slackness of the starting time of a task. We evaluated the efficacy of our algorithm using extensive simulations. Our results show that, under certain scenarios, LPST can perform 7x~70x better than the heuristics which blindly treat the infrastructure as a collection of homogeneous resources, and 46.6%~65.9% better than the algorithms that take into accountthe network topology.

The rapid adoption of smartphones with different types of advanced sensors has led to an increasing trend in the usage of mobile crowdsensing applications, e.g., to create hyper-local weather maps. However, the high energy consumption of crowdsensing, chiefly due to expensive network communication, has been found to be detrimental to the wide-spread adoption. We propose a framework, called Sense-Aid, that can provide energy-efficient mobile crowdsensing service, coexisting with the cellular network. There are two key innovations in Sense-Aid beyond prior work (Piggyback Crowdsensing-Sensys13)---the middleware running on the cellular network edge to orchestrate multiple devices present in geographical proximity to suppress redundant data collection and communication. It understands the state of each device (radio state, battery state, etc.) to decide which ones should be selected for crowdsensing activities at any point in time. It also provides a simple programming abstraction to help with the development of crowdsensing applications. We show the benefit of Sense-Aid by conducting a user study consisting of 60 students in our campus, compared to a baseline periodic data collection method and Piggyback Crowdsensing. We find that energy saving is 93.3% for Sense-Aid compared with Piggyback Crowdsensing in a representative case which requires 2 devices to provide barometric values within an area of a circle whose radius is 1 kilometer and requires periodic data collection every 5 minutes for a 90-minute test. The selection algorithm of Sense-Aid also ensures reasonable fairness in the use of the different devices.

With the explosion of data in applications all around us, erasure coded storage has emerged as an attractive alternative to replication because even with significantly lower storage overhead, they provide better reliability against data loss. Reed-Solomon code is the most widely used erasure code because it provides maximum reliability for a given storage overhead and is flexible in the choice of coding parameters that determine the achievable reliability. However, reconstruction time for unavailable data becomes prohibitively long mainly because of network bottlenecks. Some proposed solutions either use additional storage or limit the coding parameters that can be used. In this paper, we propose a novel distributed reconstruction technique, called Partial Parallel Repair (PPR), which divides the reconstruction operation to small partial operations and schedules them on multiple nodes already involved in the data reconstruction. Then a distributed protocol progressively combines these partial results to reconstruct the unavailable data blocks and this technique reduces the network pressure. Theoretically, our technique can complete the network transfer in ⌈(log2(k + 1))⌉ time, compared to k time needed for a (k, m) Reed-Solomon code. Our experiments show that PPR reduces repair time and degraded read time significantly. Moreover, our technique is compatible with existing erasure codes and does not require any additional storage overhead. We demonstrate this by overlaying PPR on top of two prior schemes, Local Reconstruction Code and Rotated Reed-Solomon code, to gain additional savings in reconstruction time.

Many storage systems are undergoing a significant shift from dedicated appliance-based model to software-defined storage (SDS) because the latter is flexible, scalable and cost-effective for modern workloads. However, it is challenging to provide a reliable guarantee of end-to-end performance in SDS due to complex software stack, time-varying workload and performance interference among tenants. Therefore, modeling and monitoring the performance of storage systems is critical for ensuring reliable QoS guarantees. Existing approaches such as performance benchmarking and analytical modeling are inadequate because they are not efficient in exploring large configuration space, and cannot support elastic operations and diverse storage services in SDS. This paper presents Inside-Out, an automatic model building tool that creates accurate performance models for distributed storage services. Inside-Out is a black-box approach. It builds high-level performance models by applying machine learning techniques to low-level system performance metrics collected from individual components of the distributed SDS system. Inside-Out uses a two-level learning method that combines two machine learning models to automatically filter irrelevant features, boost prediction accuracy and yield consistent prediction. Our in-depth evaluation shows that Inside-Out is a robust solution that enables SDS to predict end-to-end performance even in challenging conditions, e.g., changes in workload, storage configuration, available cloud resources, size of the distributed storage service, and amount of interference due to multi-tenants. Our experiments show that Inside-Out can predict end-to-end performance with 91.1% accuracy on average. Its prediction accuracy is consistent across diverse storage environments.

Multimedia streaming is a major mobile application, accounting for more than half of total mobile traffic. Streaming applications usually have a static buffering strategy. For example, buffer size is limited to x minutes of the stream, where x is optimized to provide the best trade-off between minimizing stalls and limiting waste of user's bandwidth and energy resulting from user abandonment. We show that such strategies based on information available on the mobile device alone do not work well when network conditions change dynamically, e.g., connectivity degrades due to congestion. We propose an alternative strategy using the framework called TANGO, based on a novel idea of cooperation between cellular network and mobile devices. By monitoring real-time network conditions and continuously predicting user location, our system is able to predict connectivity degradation in the near term. In such events, a notification is sent to the mobile device so that the streaming application can initiate a mitigation action, such as to pre-cache more content. In simulations based on real user traces, we found that TANGO reduces pause time by 13-72%, significantly outperforming DASH, which is the current state of the art.

A complex cloud application consists of virtual machines (VMs) running software such as web servers and load balancers, storage in the form of disk volumes, and network connections that enable communication between VMs and between VMs and disk volumes. The application is also associated with various requirements, including not only quantities such as the sizes of the VMs and disk volumes, but also quality of service (QoS) attributes such as throughput, latency, and reliability. This paper presents Ostro, an Open Stack-based scheduler that optimizes the utilization of data center resources, while satisfying the requirements of the cloud applications. The novelty of the approach realized by Ostro is that it makes holistic placement decisions, in which all the requirements of an application -- described using an application topology abstraction -- are considered jointly. Specific placement algorithms for application topologies are described including an estimate-based greedy algorithm and a time-bounded A algorithm. These algorithms can deal with complex topologies that have heterogeneous resource requirements, while still being scalable enough to handle the placement of hundreds of VMs and volumes across several thousands of host servers. The approach is evaluated using both extensive simulations and realistic experiments. These results show that Ostro significantly improves resource utilization when compared with naive approaches.

The explosive growth of mobile data traffic poses severe pressure on cellular providers to better manage their finite spectrum. Proposed solutions such as congestion-pricing exist, but they degrade users' ability to use the network when they want. In this paper, we propose a fundamentally different approach - rather than reducing the aggregate busy hour traffic, we seek to smooth the peaks that cause congestion. Our approach is based on two key insights obtained from traffic traces of a large cellular provider. First, mobile traffic demonstrates high short-term variation so that delaying traffic for very short periods of time can significantly reduce peaks. Second, by making collaborative decisions on which traffic gets delayed and by how much across all users of a cell, the delays need not result in any degradation of user experience. We design a system, CoAST, to implement this approach using three key mechanisms: a protocol to allow mobile applications and providers to exchange traffic information, an incentive mechanism to incentivize mobile applications to collaboratively delay traffic at the right time, and mechanisms to delay application traffic. We provide extensive evaluations that show that CoAST reduces traffic peaks by up to 50% even for applications that are not thought to be delay-tolerant, e.g., streaming and web browsing, but which together account for 70% of all cellular traffic.

We propose a system that makes opportunistic use of mobile computing devices and ad hoc networking to provide a transient storage service to clients in a localized geographical region. The main challenge is to offset the potential data loss caused by node mobility with internode communication. We first argue on the basis of simulation and theory that such a service is feasible, given a sufficiently high density of mobile devices. A distributed communication and storage protocol is then presented for situations where all mobile devices are within communication range of each other, and it is shown through testbed experiments and simulation that the protocol operates correctly and makes efficient use of storage space and communication bandwidth, while maximizing the longevity of stored data.

Energy conservation and reliability of wireless communications are two crucial requirements of practical sensor networks. Radio duty cycling is a widely used mechanism to reduce energy consumption of sensor devices and to increase the lifetime of the network. A side effect of radio duty cycling is that it can cause the wireless communications to be unreliable---if a sender node transmits a packet while the receiver is asleep, the communication fails. Early duty cycling protocols like B-MAC that were designed for bit streaming radios achieve low duty cycle by keeping the radio transceiver awake for short time periods. However, they require a transmitter node to precede a packet transmission with a long preamble to ensure the reliability of wireless communication. Furthermore, they cannot be used with modern packet radios like widely used IEEE 802.15.4 based radio transceivers, which cannot transmit arbitrarily long preambles. Recent duty cycling schemes like X-MAC, on the other hand, reduce the length of the preamble and are designed to work with packet radios. However, in order to ensure that a receiver can reliably detect a transmitter's preamble transmission, these schemes need to turn the radio transceiver on for longer time durations than the early schemes like B-MAC. In this paper, we present a novel duty cycling scheme called Quick MAC, that achieves a very low duty cycle without compromising the reliability of wireless communication. Furthermore, Quick MAC is stateless, compatible with packet (and bit stream) radios, and does not require synchronization among sensor nodes. From our experiments using TMote sky motes, we show that Quick MAC reduces duty cycle by a factor of about 4 compared to X-MAC, and yet maintains the same level of reliability of wireless communication as X-MAC.

Wireless reprogramming of sensor nodes is an essential requirement for long-lived networks because software functionality needs to be changed over time. The amount of information that needs to be wirelessly transmitted during reprogramming should be minimized to reduce reprogramming time and energy. In this paper, we present a multihop incremental reprogramming system called Hermes that transfers over the network the delta between the old and new software and lets the sensor nodes rebuild the new software using the received delta and the old software. It reduces the delta by using techniques to mitigate the effects of function and global variable shifts caused by the software modifications. Then it compares the binary images at the byte level with a method to create a small delta that needs to be sent over the wireless network to all the nodes. For the wide range of software change scenarios that we experimented with, we find that Hermes transfers up to 201 times less information than Deluge, the standard reprogramming system for TinyOS, and 64 times less than an existing incremental reprogramming system by Jeong and Culler.

An important but relatively less studied class of network layer protocol for sensor networks is geocast. It allows a sensor node to send messages to all nodes in a given geographical area without the sender node having any knowledge about which nodes are present in that area. Developing a robust geocast protocol for practical sensor networks poses several challenges. Geocast messages should be reliably delivered to the destination area in the presence of unreliable wireless links, a typical characteristic of practical sensor network deployments. The protocol should minimize the number of radio transmissions and avoid control traffic to save energy, which is a scarce resource in sensor networks. The protocol should be robust against a wide range of network densities. This paper presents the design, implementation, and evaluation of SGcast - a reliable, robust, and energy-efficient geocast protocol that achieves these goals. For a wide range of experiments conducted using networks of real sensor nodes and simulations, we show that compared to a recent geocast protocol, SGcast achieves up to 11.08x reduction in energy consumption and up to 2.17x improvement in successful delivery of geocast messages to the destination area, while being robust against a wide variability in network densities.

Wireless reprogramming of sensor nodes is an essential requirement for long-lived networks since software functionality needs to be changed over time. During reprogramming, the number of radio transmissions should be minimized, since reprogramming time and energy depend chiefly on the number of radio transmissions. In this article, we present a multihop incremental reprogramming protocol called Zephyr that transfers the delta between old and new software versions, and lets the sensor nodes rebuild the new software using the received delta and the old software. Zephyr reduces the delta size by using application-level modifications to mitigate the effects of function shifts. Then it compares the two binary images at the byte level to generate a small delta, that is then sent over the wireless network to all the nodes. For the wide range of software change cases that we used as benchmarks, Zephyr transfers 1.83 to 1987 times less traffic through the network than Deluge (the standard nonincremental reprogramming protocol for TinyOS) and 1.14 to 49 times less traffic than an existing incremental reprogramming protocol by Jeong and Culler [2004].

Data aggregation in wireless sensor networks is vulnerable to security attacks and natural failures. A few nodes can drastically alter the result of the aggregation by reporting erroneous data. In this paper we present RDAS, a robust data aggregation protocol that uses a reputation based approach to identify and isolate malicious nodes in a sensor network. RDAS is based on a hierarchical clustering arrangement of nodes, where a cluster head analyzes data from the cluster nodes to determine the location of an event. It uses the redundancy of multiple nodes sensing an event to determine what data should have been reported by each node. Nodes form part of a distributed reputation system, where they share information about other node's performance in reporting accurate data and use the reputation ratings to suppress reports from malicious nodes. RDAS is able to perform accurate data aggregation in the presence of individually malicious and colluding nodes, as well as nodes that try to compromise the integrity of the reputation system by lying about other nodes' behavior. We show that RDAS is more resilient to security attacks with respect to accuracy of event localization than the baseline data aggregation protocol with no security feature.

Because of transient wireless link failures, incremental node deployment, and node mobility, existing information dissemination protocols used in wireless ad-hoc and sensor networks cause nodes to periodically broadcast "advertisement" containing the version of their current data item even in the "steady state" when no dissemination is being done. This is to ensure that all nodes in the network are up-to-date. This causes a continuous energy expenditure during the steady state, which is by far the dominant part of a network's lifetime. In this paper, we present a protocol called Varuna which incurs a constant energy cost, independent of the duration of the steady state. In Varuna, nodes monitor the traffic pattern of the neighboring nodes to decide when an advertisement is necessary. Using testbed experiments and simulations, we show that Varuna achieves several orders of magnitude energy savings compared to Trickle, the existing standard for dissemination in sensor networks, at the expense of a reasonable amount of memory for state maintenance.

Wireless reprogramming of sensor nodes is a requirement for long-lived networks due to changes in the functionality of the software running on the nodes. The amount of information that needs to be wirelessly transmitted during reprogramming should be minimized to reduce reprogramming time and energy. In this paper, we present a multi-hop incremental reprogramming protocol called Hermes that transfers over the network the delta between the old and new software and lets the sensor nodes rebuild the new software using the received delta and the old software. It reduces the delta by using techniques to mitigate the effects of function and global variable shifts caused by the software modifications. Then it compares the binary images at the byte level with a method to create small delta. For a wide range of software change scenarios that we experimented with, we find that Hermes transfers up to 201 times less information than Deluge, the standard reprogramming protocol for TinyOS and 64 times less than an existing incremental reprogramming protocol by Jeong and Culler.

A specific application for wastewater monitoring and actuation, called CSOnet, deployed city-wide in a mid-sized US city, South Bend, Indiana, posed some challenges to a time synchronization protocol. The nodes in CSOnet have a low duty cycle (2% in current deployment) and use an external clock, called the Real Time Clock (RTC), for triggering the sleep and the wake-up. The RTC has a very low drift (2 ppm) over the wide range of temperature fluctuations that the CSOnet nodes have, while having a low power consumption (0.66 mW). However, these clocks will still have to be synchronized occasionally during the long lifetime of the CSOnet nodes and this was the problem we confronted with our time synchronization protocol. The RTC to fit within the power and the cost constraints makes the tradeoff of having a coarse time granularity of only 1 second. Therefore, it is not sufficient to synchronize the RTC itself---that would mean a synchronization error of up to 1 second would be possible even with a perfect synchronization protocol. This would be unacceptable for the low duty cycle operation---each node stays awake for only 6 seconds in a 5 minute time window. This was the first of three challenges for time synchronization. The second challenge is that the synchronization has to be extremely fast since ideally the entire network should be synchronized during the 6 second wake-up period. Third, the long range radio used for the metropolitan-scale CSOnet does not make its radio stack software available, as is seen with several other radios for long-range ISM band RF communication. Therefore, a common technique for time synchronization---MAC layer time-stamping---cannot be used. Additionally, MAC layer time-stamping is known to be problematic with high speed radios (even at 250 kbps).We solve these challenges and design a synchronization protocol called Harmonia. It has three design innovations. First, it uses the finely granular microcontroller clock to achieve synchronization of the RTC, such that the synchronization error, despite the coarse granularity of the RTC, is in the microsecond range. Second, Harmonia pipelines the synchronization messages through the network resulting in fast synchronization of the entire network. Third, Harmonia provides failure handling for transient node and link failures such that the network is not overburdened with synchronization messages and the recovery is done locally. We evaluate Harmonia on CSOnet nodes and compare the two metrics of synchronization error and synchronization speed with FTSP. It performs slightly worse in the former and significantly better in the latter.

As sensor networks operate over long periods of deployment in difficult to reach places, their requirements may change or new code may need to be uploaded to them. The current state-of-the-art protocols (Deluge and MNP) for network reprogramming perform the code dissemination in a multihop manner using a three-way handshake where metadata is exchanged prior to code exchange to suppress redundant transmissions. The code image is also pipelined through the network at the granularity of pages. In this article we propose a protocol called Freshet for optimizing the energy for code upload and speeding up the dissemination if multiple sources of code are available. The energy optimization is achieved by equipping each node with limited nonlocal topology information which it uses to determine the time when it can go to sleep since code is not being distributed in its vicinity. The protocol to handle multiple sources provides a loose coupling of nodes to a source and disseminates code in waves each originating at a source with a mechanism to handle collisions when the waves meet. The protocol's performance with respect to reliability, delay, and energy consumed is demonstrated through analysis, simulation, and implementation on the Berkeley mote platform.

Wireless reprogramming of a sensor network is useful for uploading new code or for changing the functionality of existing code. Through the process, a node should remain receptive to future code updates because reprogramming may be done multiple times during the node's lifetime. Existing reprogramming protocols, such as Deluge, achieve this by bundling the reprogramming protocol and the application as one program image, thereby increasing the overall size of the image which is transferred through the network. This increases both time and energy required for network reprogramming. We present a protocol called Stream that mitigates the problem by significantly reducing the size of the program image. Using the facility of having multiple code images on a node and switching between them, Stream pre-installs the reprogramming protocol as one image and the application program equipped with the ability to listen to new code updates as the second image. For a sample application, Stream reduces the size of the program image by 10 pages (48 packets/page) compared to Deluge. Stream is implemented on the Mica2 nodes and we conduct testbed and simulation experiments to show the reduction in energy and reprogramming time of Stream compared to Deluge.

Previous studies on the effect of the clipping noise on the error performance of orthogonal frequency-division multiplexing (OFDM) systems in frequency selective fading channels provide pessimistic results. They do not consider the effect of channel fading on the clipping noise. The clipping noise is added at the transmitter and hence fades with the signal. Here, the authors show that the "bad" subcarriers that dominate the error performance of the OFDM system are least affected by the clipping noise and, as a result, the degradation in the error performance of OFDM system in fading channels is very small.