Intelligent Transportation Systems (ITS) is pivotal to smart city development, relying on Internet of Things (IoT) sensors deployed along roadways. In ITS, traffic flow prediction plays a crucial role. However, the complex spatio-temporal patterns in traffic data lead to information redundancy, which increases model complexity and limits prediction performance. Existing prediction methods, ranging from statistical models to deep learning approaches, typically process such correlated spatial and temporal features indiscriminately, resulting in and limited performance gains. To address this issue, this paper proposes a novel Coupled Adaptive Sparse Graph Convolutional Network (CASGCN) for traffic prediction. The core of the method is a coupled adaptive sparse strategy that simultaneously sparsifies the input features (node features and edge features) and GCN weights to improve the learning efficiency. Specifically, we design a bi-view adaptive sparse input features module that employs spatio-temporal projection and edge construction methods to selectively preserve informative features and essential connections. Meanwhile, we propose an adaptive sparse weighting scheme for the GCN, which incorporates orthogonal constraints to promote independence across convolutional filters and effectively mitigate redundancy. Extensive experiments on four real-world traffic datasets demonstrate that CASGCN achieves competitive performance compared to state-of-the-art methods, validating the effectiveness of our framework in improving traffic prediction.

Due to the increasingly data-intensive nature of the applications, the storage system performance continues to increase in importance and is often substantially responsible for the overall processing rate of the application. Fortunately, the storage technologies themselves are improving rapidly in numerous ways, from low-level read/write of bits in a device all the way to the management of the entire storage hierarchy in large enterprise and cloud settings. Studying many of the important issues in this entire spectrum often requires storage access traces from the storage server side, but these are often hard to come by. To address this gap, we present a method to generate synthetic traces using a novel generative adversarial network (GAN) architecture that captures the realism and diversity of real storage traces. The generated traces can be used to augment the existing workload traces of interest for a variety of storage system studies. We demonstrate how the proposed method can generate storage traces that have the overall characteristics of the real traces and yet provide behavioral diversity.

In this paper we explore a Partial Digital Twin (PDT) framework for online diagnosis of faults in Microservices ($\mu S$ s) based systems. The microservices environment is highly dynamic and suffers from configuration problems due to frequent changes driven by CI/CD; therefore, quick online diagnosis is crucial but has not been addressed in the literature. We have designed the following mechanisms: (1) a conversational LLM assisted interface to automatically generate the initial “troubleticket” based on the problem(s) encountered by the user, (2) integration of tests for several key $\mu S$ s patterns to diagnose distributed transaction failures, and (3) ensuring robustness of fault diagnosis scheme. Our methodology leverages the service mesh capabilities for $\mu S$ s to dynamically reconstruct transactional contexts while incorporating a hierarchical testing strategy. Experimental results demonstrate that the framework requires only $8 \%$ more tests on average beyond the theoretical optimal for complex transaction chains. The LLM component shows an initial fault categorization accuracy of $92 \%$, enabling precise reproduction of failure scenarios in the PDT environment without disrupting the production system.

In this paper we explore strategies for providing energy to nodes of small intrabody networks of sensors, actuators, and decision support nodes that are deployed to manage chronic illnesses. In particular, we study both a centralized and decentralized mechanism for energy transfer by using magnetic resonance communication (MRC), which is suitable for through-the-body communications up to a meter distance. We demonstrate that while the centralized mechanism can maintain a higher energy level for the critical "hub" nodes of the network, and decentralized mechanism not only has a lower overhead but is also more robust and can deal with adverse events (e.g., energy starvation) better. We also explore a hybrid strategy in order to harness the advantages of both centralized and decentralized schemes.

The increasing adoption of DevOps and microservices paradigm in enterprise software has not only brought about several benefits but also a host of challenges. In particular, such applications may suffer from frequent misconfigurations and other faults that need to be diagnosed promptly. In this paper, we propose a novel framework centered around the concept of Partial Digital Twin (PDT) to enable isolated testing without disrupting the main infrastructure. Our framework employs a systematic methodology to replicate the chain of dependencies relevant to reported faults, facilitating efficient root cause analysis. Additionally, we introduce a hierarchical categorization of faults leading to the selection of tests tailored to specific fault types. Through empirical evaluation, we demonstrate that the median number of tests required to diagnose a misconfiguration using our approach is only 7% above the ideal scenario.

This paper explores the application of Magnetic Resonance Coupling (MRC) and Ultrasonic Coupling (USC) for through-the-body wireless communication (TBWC) for biomedical applications. Detailed (on-body/in-body) simulations using the Sim4Life package and real-world on-body experiments are used to evaluate signal propagation, power transfer efficiency, and path-loss characteristics through this complex media. The results indicate that although USC has a somewhat smaller attenuation at lower frequencies, MRC demonstrates a more consistent performance with frequency variation and a smaller difference between in-body and on-body scenarios. These results, coupled with the fact that the lowest attenuation occurs around 25–30 MHz, for which USC transducers are very difficult to design, suggests that MRC provides a desired technology for intrabody networks.

The deployment of Internet of Things (IoT) systems spans a large variety of applications, each with unique requirements. Many of these applications relate to the management of various cyber–physical systems, including road and rail traffic, electricity, water, food/goods transportation and storage, smart building management, crime/safety management, underwater systems, etc. Within this context, a growing concern is the deployment of IoT systems in extreme environments which may occur either because of the nature of the application or due to external factors. Some prominent examples of the former are 1) the IoT deployments in hazardous environments such as management of chemical or nuclear plants, management of underwater oil/gas infrastructure, mining operations, etc.; 2) IoT systems deployed specifically to manage accidents and disasters; and 3) IoT systems deployed in arctic/antarctic regions where they routinely experience extreme levels of changes in terms of temperature, wind conditions, sunlight availability, compression, etc. Such IoT systems generally are designed to specifically operate in the challenging environment they must operate in, and thus may be expected to be rather robust. However, the IoT systems designed to manage the physical infrastructures such as those in urban settings may also need to handle unprecedented and unexpected stresses due to the worldwide phenomena of aging physical infrastructure, demand that far exceeds the designed capacity, and increasingly extreme operating conditions due to climate change. This special issue covers all such scenarios and thus represents a vast and rich area for innovations.

In this article, we demonstrate that deep learning (DL) and spatiotemporal reasoning can effectively identify driving behavior based on the videos captured by roadside cameras. The use of roadside infrastructure for such determination is twofold: 1) a global view of the vehicles and their interactions and 2) no involvement or awareness of the vehicles or their drivers, so the determination is inexpensive, easy to deploy, and entirely nonintrusive. Furthermore, our method uses DL only for object detection and tracking and builds a flexible and explainable reasoning model to identify the driving behavior. The essential advantage of this approach is that we use DL only for tasks that can be accomplished efficiently and with high accuracy (i.e., object detection and tracking), which can be done in real time. Although there are DL models for detecting complex activities (e.g., aggressive driving), they are much harder to train, require higher accuracy, and inferencing time may not satisfy real-time constraints. By using a setup with program-controlled robocars, we demonstrate that we can achieve accuracies of 98% and 99% for driving behavior characterization, and the mechanism can provide detection of 650 ms on a very dated desktop. The characterization can provide feedback to the driver (or the automated car) for improved traffic safety and roadway throughput.

Federated learning (FL) is a promising distributed machine learning scheme where multiple clients collaborate by sharing a common learning model while maintaining their private data locally. It can be applied to a lot of applications, e.g., training an automatic driving system by the perception of multiple vehicles. However, some clients may join the training system dynamically, which affects the stability and accuracy of the learning system a lot. Meanwhile, data heterogeneity in the FL system exacerbates the above problem further due to imbalanced data distribution. To solve the above problems, we propose a novel FL framework named FedREM (Retain-Expansion and Matching), which guides clients training models by two mechanisms. They are 1) a Retain-Expansion mechanism that can let clients perform local training and extract data characteristics automatically during the training; 2) a Matching mechanism that can ensure new clients quickly adapt to the global model based on matching their data characteristics and adjusting the model accordingly. Results of extensive experiments verify that our FedREM outperforms various baselines in terms of model accuracy, communication efficiency, and system robustness.

Given the difficulty in obtaining adequate data from production systems, characterizing performance as a function of configuration variables (CVs) via supervised learning is difficult, and the use of standard semi-supervised learning (SSL) techniques may or may not help. In this paper, we describe a knowledge-assisted (KA) SSL algorithm that determines the confidence level of the generated data independently based on the domain knowledge. We demonstrate that such an approach outperforms plain SSL with the most popular SSL algorithms for all the workloads used in this study.

I have the great pleasure to welcome you to the International Conference on Computer Communications and Networks (ICCCN 2023) which takes place July 24−26, 2023. It has been an honor to serve as the General Chair of the 32nd edition of the conference.

This special session will report on the updated NSF/IEEE-TCPP Curriculum on Parallel and Distributed Computing released in Nov 2020 by the Center for Parallel and Distributed Computing Curriculum Development and Educational Resources (CDER). The purpose of the special session is to obtain SIGCSE community feedback on this curriculum in a highly interactive manner employing the hybrid modality and supported by a full-time CDER booth for the duration of SIGCSE. In this era of big data, cloud, and multi- and many-core systems, it is essential that the computer science (CS) and computer engineering (CE) graduates have basic skills in parallel and distributed computing (PDC). The topics are primarily organized into the areas of architecture, programming, and algorithms topics. A set of pervasive concepts that percolate across area boundaries are also identified. Version 1 of this curriculum was released in December 2012. That curriculum guideline has over 140 early adopter institutions worldwide and has been incorporated into the 2013 ACM/IEEE Computer Science curricula. This Version-II represents a major revision. The updates have focused on enhancing coverage related to the topical aspects of Big Data, Energy, and Distributed Computing.

Situational awareness tries to grasp the important events and circumstances in the physical world through sensing, communication, and reasoning. Tracking the evolution of changing situations is an essential part of this awareness and is crucial for providing appropriate resources and help during disasters. Social media, particularly Twitter, is playing an increasing role in this process in recent years. However, extracting intelligence from the available data involves several challenges, including (a) filtering out large amounts of irrelevant data, (b) fusion of heterogeneous data generated by the social media and other sources, and (c) working with partially geo-tagged social media data in order to deduce the needs of the affected people. Spatio-temporal analysis of the data plays a key role in understanding the situation, but is available only sparsely because only a small fraction of people post relevant text and of those very few enable location tracking. In this paper, we provide a comprehensive survey on data analytics to assess situational awareness from social media big data.

In this paper, we present C-FAR, a framework for reasoning about anomalies in road-based intelligent transportation systems (ITS) based on video monitoring by the roadside camera infrastructure. The anomalies could span broad temporal and spatial ranges, including fine-grain (e.g., unsafe interactions among moving vehicles in real-time), medium-grain (e.g., aggressive/unsafe driving styles of individual vehicles over extended periods/distances), and coarse-grain (e.g., ensemble properties of the traffic over even longer time horizons). Unlike traditional approaches that utilize deep learning to recognize individual activities, C-FAR does so only for primitive movements and activities and then builds a comprehensive event logic framework. It also provides an optimal resolution of the detected/predicted anomalies by identifying the minimal changes in the controllable parameters of the system. We implemented a prototype system and tested it on three distinct real-world traffic data sets. We demonstrate that the proposed scheme can predict anomalies with over 84% recall level at 95% confidence level approximately 4.05 seconds before the incident.

Distributed machine learning (ML) was originally introduced to solve a complex ML problem in a parallel way for more efficient usage of computation resources. In recent years, such learning has been extended to satisfy other objectives, namely, performing learning in situ on the training data at multiple locations and keeping the training datasets private while still allowing sharing of the model. However, these objectives have led to considerable research on the vulnerabilities of distributed learning both in terms of privacy concerns of the training data and the robustness of the learned overall model due to bad or maliciously crafted training data. This article provides a comprehensive survey of various privacy, security, and robustness issues in distributed ML.

The increasing use of the DevOps paradigm in software systems has substantially increased the frequency of configuration parameter setting changes. Ensuring the correctness of such settings is generally a very challenging problem due to the complex interdependencies, and calls for an automated mechanism that can both run quickly and provide accurate settings. In this paper, we propose an efficient discrete combinatorial optimization technique that makes two unique contributions: (a) an improved and extended metaheuristic that exploits the application domain knowledge for fast convergence, and (b) the development and quantification of a discrete version of the classical tunneling mechanism to improve the accuracy of the solution. Our extensive evaluation using available workload traces that do include configuration information shows that the proposed technique can provide a lower-cost solution (by ~60%) with faster convergence (by ~48%) as compared to the traditional metaheuristic algorithms. Also, our solution succeeds in finding a feasible solution in approximately 30% more cases than the baseline algorithm.

Large-scale loT systems are likely to involve mul-tiple subsystems deployed and operated by different “parties”, which must collaborate to ensure that their operational rules do not conflict. We codify the smooth functioning of the entire system through a set of “safety properties” that must be enforced collaboratively. However, this requires cross-party access to the sensors/actuators state and the ability to request remote actuations. In this paper, we define an access control architecture for such situations where we distinguish between the static authorization problem that selects parties tasked with safety property enforcement and the dynamic (run-time) control over accesses. This results in a unique enforcer selection problem for which we develop efficient algorithms and quantify their performance through a comprehensive emulation of an extensive smart home. We also show that the additional cost of granting access rights to the parties is quite small in a medium-size emulated multiparty loT environment.

Biomedical systems of implanted miniaturized sensors and actuators interconnected into an intra-body area net-work could revolutionize treatment options for chronic diseases afflicting internal organs. Considering the well-understood limitations of radio frequency (RF) propagation in the human body, we have explored magnetic resonance (MR) coupling for both communications and energy transfer through the body. In this paper, we have discussed the design and implementation of a software-defined prototype using Universal Software Radio Peripheral (USRP) boards. We have reported experimental results on the achieved packet error rates at different positions through-the-body distances and packet sizes. We have observed experimentally that the MR signal propagates through the body substantially better than in the air, and can provide a practical means for energy transfer and communications in intra-body networks. It also works better than the better understood galvanic coupling.

Human body communication (HBC) has recently been explored extensively both for small wearable electronic gadgets and for implanted sensors to deliver relevant data to implanted therapeutic devices. In this paper, we conduct an experimental comparison of two of the promising technologies but for on-body use, namely ultrasound coupling (USC) and magnetic resonance coupling (MRC) based communications. We find that both of these propagate much better through the body than in the air, thereby making them attractive for communications between in-body nodes, in-body to on-body nodes, and on-body nodes where the direct path includes substantial body area. USC also involves a surface acoustic wave (SAW) between on-body nodes which may be broken to varying extent by clothing. We find that with SAW component, USC works better than MRC, but otherwise has similar performance. MRC is very robust and can travel up to the entire body length with 25dB or less loss.

On behalf of the Technical Program Committee, we welcome you to the International Conference on Computer Communications and Networks (ICCCN), 2022. This year we celebrate the 31st anniversary of the conference, and continuing its tradition, we aspired in delivering an exciting and of high quality technical program, while aiming to bring together researchers, designers, and implementers of all aspects of computer communications and networks.

In this paper, we address the characterization of similarity between two storage traces and define a three-part measure called SIST (Similarity Index for Storage Traffic). Such a measure is essential for identifying traces that are most appropriate for storage system evaluations. We compare SIST against several other similarity measures in the literature on both the object storage and block storage systems, and show the superiority of SIST in terms of its behavior for known perturbations.

The emerging systems are not only generating huge amounts of data but also expect this data to be analyzed expeditiously to drive online decision-making and control. Thus, identifying the most relevant data and making it available close to the computation becomes a central challenge in driving the big data revolution. Storage systems play a crucial role in enabling efficient access to the stored data and intelligent storage management techniques are thus central to addressing the problem. Generally, as the data volume increases, the marginal utility of an “average” data item tends to decline, which requires greater effort in identifying the most valuable data items and making them available with minimal overhead and latency. Data driven mechanisms have a big role to play in solving this needle-in-the-haystack problem. In this paper we propose a taxonomy to provide a structure for understanding the common issues surrounding these techniques. We discuss these techniques and articulate many research challenges and opportunities.

The emerging edge computing applications often use high definition cameras as edge devices to capture video streams that need to be analyzed in real-time for situational understanding and answering queries. However, such devices suffer from limited energy (and hence limited computing power) and limited bandwidth available to stream the data to the edge controllers that provide much higher computing capacities. In this paper, we address these issues in the context of vehicular traffic monitoring and develop a scheme that has two components: YLLO and BATS. YLLO is a lightweight object recognition algorithm that runs on the edge device itself and substantially reduces the frame rate sent to the edge controller without removing the important information. BATS adapts the transmissions to the available bandwidth by taking advantage of further redundancy in the video stream in both single and multi-camera scenarios. We show that these mechanisms together can maintain object identification accuracy of above 95 percent, while transmitting just$\sim$5–10 percent of all the frames recorded by the cameras.

In this paper we examine the problem of conflict detection and mitigation across multiple independently designed IoT subsystems deployed in a shared environment. The desired behavior of the system is codified in terms of predefined "safety properties". We allow both the operational rules and safety properties to include time and temporal logic operations and detect their potential violation proactively via a "look ahead" mechanism. The problematic operational rules are then perturbed within the allowable range for mitigation. We show that our mitigation approach, based on intelligent combinatorial optimization, can resolve the conflicts via perturbation in 100% of the cases where such a resolution is feasible.

In this paper, we propose a quality of service (QoS) aware transport solution for storage access over data center network. Motivation for QoS differentiation comes from the emerging storage technologies which not only provide network comparable latency but also overwhelm the network bandwidth close to the storage servers. We consider both throughput and latency related QoS requirements and demonstrate how they can be enforced by Explicit Congestion Notification (ECN) enabled switches. The mechanism can be viewed as a way of adding QoS capability to the existing data center transport solutions. Our scheme can co-exist with existing congestion management schemes and ensures RTT fairness while providing service differentiation. We show that our scheme can provide differentiated treatment and besides achieving better throughput fairness during the congestion episode, our solution can reduce the target latency misses by up to 71% and 80% as compared to the existing TCP and RDMA transport.

Solid-State Drives (SSDs) involve a complex set of management activities in the background, resulting in unpredictable delays and occasional extended access latencies. However, there is an increasing demand for "deterministic" access latency in a growing number of scenarios. This demand has prompted a new feature in the NVMe storage access protocol called Predictable Latency Mode (PLM), which provides a way to tighten tail latency in SSDs. This paper presents the first study of the PLM feature in a single-host environment and its extension to multi-host settings. We propose a PLM Coordinator (PLMC) that regulates access to the PLM of a shared SSD device based on the hosts’ traffic characteristics. Our simulation experiments show that the proposed PLMC can achieve 82% improvement in 99.99% tail latency compared to a bare SSD without PLM feature. Moreover, the proposed coordinator with simple traffic prediction can perform 93.2% better than without coordinator on the 99%-tail latency values.

The interactive web applications increasingly demand an end-to-end latency that is not only low on the average but also is “deterministic” in that they avoid long tails. Storage systems today largely keep data in SSDs, but SSDs are known to have unpredictable latencies due to background activities such as garbage collection. The recent NVMe access protocol proposes a Predictable Latency Mode (PLM) which allows the SSD to cycle between deterministic window (DTWin) and nondeterministic window (NDWin) periods, with background activities largely pushed to the latter. However, this means that the number of read and write IOs during DTWin period is limited and need to be managed properly. Another challenge is that to date no real SSDs are available in the market with this feature. In this paper, we explore the possibility of emulating the PLM feature in regular SSDs using Intel Optane that does provide a rather deterministic access latency. In particular, we propose a write I/O friendly PLMlightcoordinator (PLMLC) that buffers writes in Optane and sends them to SSD during the NDWin-like period and also intelligently manages the limited number of IOs possible during DTWin-like period. The coordinator is designed to handle requests from multiple hosts that may access shared data on the SSD and may have different QoS requirements in terms of latencies. The results show that PLMLC improves the 99%-ile tail latency by 5.8x even without any sophisticated traffic estimation procedures.

Blockchain technology has gained a significant amount of interest in recent years due to its decentralized control, immutability, transparency and robustness. In this paper we propose an enhancement to BlockChain built using proof-of-elapsed-time (PoET) consensus protocol to further increase its efficiency and transaction throughput. The proposed scheme, called DC-PoET, exploits distributed coordination (DC) among the nodes to avoid unnecessary transmission of conflicting blocks inspired by a similar mechanism in WiFi networks. We show that DC-PoET can support around 465 transactions per seconds with 30 MB block size, and even higher for larger blocks. We have also developed detailed analytical modeling for the performance of DC-PoET scheme using a two-dimensional Markov Chain, along with the validation of such modeling using Matlab simulations. The security analysis of our proposed scheme is also discussed.

The papers in this special section focus on configuration security for industrial automation and control systems. These systems include supervisory control and data acquisition systems, distributed control systems, and other control system configurations such as programmable logic controllers, which are typically used in industries such as electric, water and wastewater, oil and natural gas, transportation, chemical, pharmaceutical, food and beverage, and discrete manufacturing, examples of which are automotive, aerospace, and durable goods. These systems are highly interconnected and mutually dependent in complex ways, both physically and through information and communications technologies, and they support a diverse set of services for the management of critical infrastructure by making use of a wide variety of Internet of Things (IoT) devices for sensing and actuation. These papers highlight the main research challenges and solutions for improving configuration security in the context of industrial automation and control systems by taking into consideration various challenges faced by industrial applications.

In this article, we propose a prekeying-based integrity protection mechanism for critical smart grid communications that are often left unprotected due to tight timing constraints. Our mechanism computes the key for the next message in advance followed by a simple exclusive-or operation with the message when it is generated. This provides both integrity and confidentiality at a very low latency cost. The rigorous security analysis shows that the proposed method is secure against cyclic redundancy check (CRC) and message replay attacks. The extensive evaluation shows that the method is up to 21 times faster than standard integrity protection algorithms, and can do the message encryption in under 1 ms even on a very low-end microcontroller.

The autonomous vehicular technology is approaching a level of maturity that gives confidence to end-users in many cities around the world for their usage so as to share the roads with manual vehicles. Autonomous and manual vehicles have different capabilities which may result in surprising safety, security, and resilience impacts when mixed together as a part of the intelligent transportation system (ITS). For example, autonomous vehicles can communicate electronically with one another, make fast decisions and associated actuation, and generally act deterministically. In contrast, manual vehicles cannot communicate electronically, are limited by the capabilities and slow reaction of human drivers, and may show some uncertainty and even irrationality in behavior due to the involvement of humans. At the same time, humans can react properly to more complex situations than autonomous vehicles. Unlike manual vehicles, the security of computing and communications of autonomous vehicles can be compromised thereby precluding them from achieving individual or group goals.

Content-Centric Networking (CCN) is a promising framework for the next generation Internet architecture that exploits ubiquitous in-network caching to minimize content delivery latency and reduce network traffic. In this paper, we introduce a neighborhood aware mechanism for content caching, named Neighborhood Aware Caching and Interest Dissemination (NACID) that accounts for the popularity of contents and how close the content copies are in the neighborhood. We use a very low-overhead, Bloom Filter based dissemination of caching information in the neighborhood. Given the neighborhood cached contents, the proposed scheme decides when and how to handle the additional caching of content and its eviction. Simulation results show that NACID performs substantially better than the existing CCN caching policies. We also study different heterogeneous cache memory allocation strategies and show that the simpler homogeneous allocation strategies work almost as well.

This paper explores the coupling between power/thermal aspects of data center and the vibrations caused by chassis/server fans in environments dominated by mechanical disks. We show that the leakage power and cooling fan induced vibrations could substantially degrade the storage system performance which in turn could lead to a poor performance and high energy consumption. We then propose our Environment Aware Data Control (EADC) management policy that utilizes built-in instrumentation telemetry and fan-speed control to minimize overall energy consumption while providing a better overall performance. Based on actual experimentation with storage servers in Oracle labs, we show that the proposed mechanism can reduce the energy consumption by 63% while simultaneously reducing the IO delay by 60% as compared to the baseline system with no vibration aware controls.

In this paper, we propose a novel caching mechanism, called FussyCache, that differs from the traditional DRAM caching mechanisms that automatically cache a data block when it is requested. Instead, FussyCache evaluates each requested data block for its caching eligibility, and reads ineligible blocks directly from the device each time. We show that the Fussy-Cache performs substantially better than the traditional caching algorithms and its performance increases with the storage device speed. In particular, for the first generation Intel Optane based storage, FussyCache provides 25-30% reduction in the average access latency as compared to the native caching mechanism such as plain LRU. We also observe close to 15-20% improvement in performance even for a mainstream TLC SSD. Furthermore, the FussyCache design includes two mechanisms that allow for its easy deployment in any environment: (a) a self-monitoring stage that reverts it to a normal LRU when partial caching is not beneficial, and (b) a training phase that automatically tunes the configurable parameters for the deployment environment.

Key-value stores (KVS) provide an efficient storage for increasing amounts of semi-structured or unstructured data generated by many applications. Most KVS in existence have been designed for hard-disk based storage where avoiding random accesses is crucial for good performance. Unfortunately, the resulting storage structures result in high read, write, and space amplifications when used on modern SSDs. In this paper, we introduce a KV store especially designed for SSDs, called FlashKey, and demonstrate that even as an initial implementation, it substantially outperforms the two most popular commercial KVS in existence, namely, Google's LevelDB and Facebook's RocksDB. In particular, we show that FlashKey achieves up to 85% improvement in average access latency, 2x improvement in tail latencies, and 12x improvement in write amplification, at comparable or better space-amplification. Furthermore, FlashKey can easily trade off space and write amplifications, thereby providing a new tuning knob that is difficult to implement in LevelDB and RocksDB.

Recent advances in artificial intelligence, edge computing, and big data have enabled extensive reasoning capabilities at the edge of the network. Edge servers are now capable of extracting meaningful intelligence from IoT nodes, which can benefit a very diverse set of IoT applications, including smart carrier and distribution networks (power, people, water, and food), smart agriculture and manufacturing, and healthcare and maintenance. Unfortunately, as the infrastructures become more intelligent, they also become more vulnerable to disruption due to cyberattacks and information leakage. Furthermore, the rich data gathering and analytics involved in driving the intelligent management substantially raise the stakes in terms of privacy violation of the people and organizations that it serves.

Large networks of IoT devices, each consisting of one or more sensors, are being increasingly deployed for comprehensive real-time monitoring of cyber-physical systems. Such networks form an essential component of the emerging edge computing paradigm and are expected to increase in complexity and size. The physical phenomenon sensed by different sensors (within the same or different IoT devices in close proximity) often have relationships that makes them correlated. This is a form of proxy sensing that can be exploited for achieving better energy efficiency and higher robustness in monitoring. In this article, we explore how a set of sensors can optimize its data collection rates efficiently in a semi-distributed manner and yet provide the advantages of autonomy, relative isolation, and distributed control that is essential in a large-scale network.

Electronic communication is crucial for monitoring the rescue-relief operations and providing assistance to the affected people during and after disasters. Given the ubiquity of smartphones, we envision that smartphones with lost connection (due to damage) to the communications infrastructure are nevertheless integrated seamlessly into the network as far as possible. To achieve this, we propose to build ad hoc subnetworks of disconnected smartphones using the WiFi tethering technology and ultimately connect them to either the emergency communication equipment deployed in the disaster area or to other smartphones that have still the network connectivity. The proposed architecture for such integration and a defined software-based control through the emergency control center (ECC) enables battery aware collection of critical data through smartphone sensors. The developed solution supports mobility of all smartphones, including those that have lost direct cellular connectivity as well as those that have not and are willing to act as gateways. We demonstrate how the proposed scheme can be tied to the standardized wireless emergency alert service and how it can effectively handle mobility tolerant device discovery and data transfer.

Transportation and distribution (T8D) of fresh food products is a substantial and increasing part of the economic activities throughout the world. Unfortunately, fresh food T8D not only suffers from significant spoilage and waste, but also from dismal efficiency due to tight transit timing constraints between the availability of harvested food until its delivery to the retailer. Fresh food is also easily contaminated, and together with deteriorated fresh food is responsible for much of food-borne illnesses. The logistics operations are undergoing rapid transformation on multiple fronts, including infusion of information technology in the logistics operations, automation in the physical product handling, standardization of labeling, addressing and packaging, and shared logistics operations under 3rd party logistics (3PL) and related models. In this article, we discuss how these developments can be exploited to turn fresh food logistics into an intelligent cyberphysical system driven by online monitoring and associated operational control to enhance food freshness and safety, reduce food waste, and increase T8D efficiency. Some of the issues discussed in this context are fresh food quality deterioration processes, food quality/contamination sensing technologies, communication technologies for transmitting sensed data through the challenging fresh food media, intelligent management of the T8D pipeline, and various other operational issues. The purpose of this article is to stimulate further research in this important emerging area that lies at the intersection of computing and logistics.

Spatial clustering of the events scattered over a geographical region has many important applications, including the assessment of needs of the people affected by a disaster. In this paper we consider spatial clustering of social media data (e.g., tweets) generated by smart phones in the disaster region. Our goal in this context is to find high density areas within the affected area with abundance of messages concerning specific needs that we call simply as “situations”. Unfortunately, a direct spatial clustering is not only unstable or unreliable in the presence of mobility or changing conditions but also fails to recognize the fact that the “situation” expressed by a tweet remains valid for some time beyond the time of its emission. We address this by associating a decay function with each information content and define an incremental spatial clustering algorithm (ISCA) based on the decay model. We study the performance of incremental clustering as a function of decay rate to provide insights into how it can be chosen appropriately for different situations.

Magnetic Induction (MI) based communication is a near-field communications technology that can work reliably in a variety of difficult propagation media and thus can be useful in many short-range IoT applications. In this paper, we explore low-power protocols for MI communications with low data rate requirements but a premium on energy consumption. In particular, we exploit communication through silence (CTS), and show that it can reduce the energy expenditure of the communication by up to 67% as opposed to typical binary packet based transmission. We also discuss a multi-channel tree based routing protocol to reduce energy consumption from overhearing in asynchronous MI communication networks and show that the proposed scheme can reduce the overhearing counts by ~60% with two channels and by ~80% with four channels.

The largest computing systems routinely run into silent data corruption (SDC) as part of its normal operation. The number of SDCs will increase drastically as computing systems approach the exascale mark, forcing a need to reconsider the resilience approach taken to counteract the effects of unmitigated data corruption errors. Yet any resilience method must be sensitive to both resource and energy requirements. In this paper we explore the propagation of data corruption errors caused in stencil computation, an iterative kernel with structured communication pattern that is found in a wide variety of scientific and engineering problems. We present a computational model, refered to as mimic replication, that provides resilience against SDC errors through dynamic reexecution of processes that are vulnerable to having their data tainted due to a detected latent error. We then provide an analytical model that allows tradeoff between resource and energy consumption and resilience.

The increase in data generation and the decrease in the cost of storage increases the need for intelligent data management. One avenue that would allow storage systems to manage data better is to feed it with an accurate prediction of how many disk operations are expected on that data. In this paper, we introduce a method to predict the data heat in a storage system. Our method is derived from two insights. The first is that we can use a set of quick to compute signals which constrain the set of possible future heat patterns. The underlying assumption in this is that the signals provide a description of the access pattern and the requests that have similar signals have a similar set of future behavior. The second is that the storage requests can be partitioned into groups based on their signal. Our method generates a unique prediction model for each group from the corresponding heat patterns. This makes the model generation process easier and more precise. The results of our method on public datasets show that it is a viable way to predict heat. Our method is able to accurately almost all inactive regions and provides useful predictions for active regions.

Effective control of emerging cyberphysical systems such as smart transportation, smart health-care, etc. requires edge computing infrastructure that is often organized into three layers, namely edge (IoT) devices, edge controllers (ECs) and the cloud. In large infrastructures, ECs must be deployed densely in the proximity of edge devices and need to satisfy strict constraints on cost, size, cooling, etc. Thus, ECs cannot host large amounts of local storage and instead must make use of cloud storage in the background to provide an impression of large, fast local storage to host the IoT device data needed for online and real-time queries. In this paper, we provide insights into the configuration issues of such an edge storage infrastructure (ESI) based on the evaluation of commercial ESIs on several real-world edge computing workloads. We also show that the current ESI designs are lacking in several respects, and suggest some approaches for enhancing their capabilities to meet the stringent requirements of emerging edge computing applications.

Radio frequency (RF) communications, although most popular, are unsuitable for environments involving aqueous and animal/plant tissue media, dense environments (e.g., small regions with many radios), applications requiring extremely low power consumption, etc. For such environments, magnetic induction (MI) communications are an emerging technology that appears to be very attractive. Although MI communication has been studied for some RF-challenged environments such as underwater, underground and body area networks, most of the studies so far are simulation based with minimal experimentation. In this paper, we show the feasibility of the proposed MI communications, by developing a small testbed using Freelinc boards. We also compare and contrast the existing theoretical claims regarding MI communications with some detailed experimental outcomes, and show how much they differ.

In this article, we introduce the concept of Water Flow Driven Sensor Networks for leakage and contamination monitoring in urban water distribution systems. The unique aspect of our work is that the sensor network can be deployed in the underground water network with only access to connection points (through manholes) and driven only by water harvested energy without the need for AC power or frequent battery changes. Although water systems may be affected by a large variety of contaminants, only a few sensors can be practically deployed. Thus, many types of contaminants are sensed via “proxy sensing,” which may not be 100% reliable. The main problems addressed are (a) adaptation of the network to the available energy to maximize leak/contamination detection and (b) minimal artificial water circulation or leakage to improve detectability during periods of almost zero natural water flow. The article shows, through extensive simulations, that the proposed approach can drastically reduce the leakage/contamination reporting time (from 3.5h up to ∼6min), and the adaptation can reduce this circulation by ∼33% and yet enhance the collected/transmitted data by 30%.

Emergency communication networks are crucial for monitoring and providing assistance to affected people during long-persisting disasters. Given substantial and increasing penetration of smart-phones throughout the world, we envision future emergency networks to consist of smart-phones in the disaster area, un-failed portions of the cellular network, and the communication capabilities provided by the specially deployed emergency equipment (e.g., fixed and mobile wireless access points deployed on the ground or in air via helicopters, satellite radio interfaces, etc.). We envision the emergency network for bulk data transmission - such as transmitting multiple pictures/sounds captured by the phone to help with rescue/safety assessment while keeping the delay/energy expenditure minimum. With extensive simulations, we show that the proposed scheme forwards the sensed data to the control centers with small latency (<; 3 seconds) while keeping the WiFi radios on for less than 1% of time.

Applications and network services in enterprise cloud environments have direct and indirect dependencies. The configuration of these services varies based on business needs. However, accurate and complete documentation of the configuration may not exist at all times. Thus, failure diagnosis becomes further complex with such unknown/uncertain dependencies. To cope with this, some probing stations need to be installed in suitable locations in the network to provide full monitoring and diagnosing capability. In this paper we develop a novel CloudMiner architecture for failure diagnosis in enterprise clouds that consist of developing intelligent probing station selection, failure detection and diagnosis across the network components using the minimum set of network probes, considering the inter-dependencies across the network services/components. Extensive simulation results show that CloudMiner can always identify the faulty components among the list of a small set of suspected components, the size of which is as low as ~3 for a network with 460 components.

The paper proposes a situational awareness service, named StayTuned that collects information from social media, extracts relevant messages, and broadcasts them to the subscribers through wireless emergency alert system. StayTuned uses automated filtering and summarization of messages and updates subscribers with real-time situational summaries. Extensive experiments were conducted using twitter data collected during the Sandy hurricane to evaluate performance of the automated message extraction.

Most large systems involve multiple zones with distinct ownership or control and one or more such zones must be traversed by the transactions. Therefore, in case of misconfigurations, it is necessary to conduct tests that go across parties. Such tests are complex as they must consider composition of test functionalities provided by each party and the feasible tests must abide by the access restrictions. In this paper we propose a framework for defining test functionalities their composition, and their access control. We then discuss an efficient algorithm to determine the realization of the given test via valid compositions of individual functionalities in a way to minimize the number of parties involved.

Content-Centric Networking (CCN) is a promising framework for the next generation Internet architecture, by exploiting ubiquitous in-network caching to minimize content delivery latency and reducing the network traffic. In this paper, we introduce a neighborhood aware mechanism for content caching, named Neighborhood Aware Caching and Interest Dissemination (NACID) that accounts for the popularity of contents and how close the content copies are there in the neighborhood. We have adopted a Bloom Filter based dissemination of caching information in the neighborhood so that its overhead remains small. Given the neighborhood cached contents the proposed scheme decides when and how to handle the additional caching of content and its eviction. Simulation results show that NACID provides an increase in up to ~3 times of cache hits, and decrease in up to ~30% the number of hops required to get the contents than existing CCN caching policies.

Sensing of food spoilage and contamination is an active area of research, with many types of contact and noncontact sensors are being developed that can track fresh food quality throughout the distribution process. In this paper, we consider the communication of the sensed product quality along with the box position (in the stack of boxes in the truck or in a warehouse room) to the next level, in order to make the logistics more efficient and less wasteful. Given the water-rich, inhomogeneous biological media, RF or ultrasonic based communications are inappropriate in such environments, and we instead explore Magnetic Induction (MI) based communication framework in the HF band (3-30 MHz). We propose a novel magnetic induction based localization scheme to localize the boxes and study its accuracy via extensive simulations. We show that with a small number of anchor nodes, the localization can be done without any errors for boxes as small as 0.5 meter on the side, and with small errors even for boxes half as big. Our preliminary analysis suggests that such sensors can last for several years without any battery replacement.

Due to the mission-critical nature of energy management, smart power grids are prime targets for cyber-attacks. A key security objective in the smart grid is to protect the integrity of synchronized real-time measurements taken by phasor measurement units (PMUs). The current communication protocol in substation automation allows the transmission of PMU data in absence of integrity protection for applications that strictly require low communication latency. This leaves the PMU data vulnerable to man-in-the-middle attacks. In this paper, a lightweight and secure integrity protection algorithm has been proposed to maintain the integrity of PMU data, which fills the missing integrity protection in the IEC 61850-90-5 standard, when the MAC identifier is declared 0. The rigorous security analysis proves the security of the proposed integrity protection method against ciphertext-only attacks and known/chosen plaintext attacks. A comparison with existing integrity protection methods shows that our method is much faster, and is also the only integrity protection scheme that meets the strict timing requirement. Not only the proposed method can be used in power protection applications, but it also can be used in emerging anomaly detection scenarios, where a fast integrity check coupled with low latency communications is used for multiple rounds of message exchanges.

In this paper, we discuss the issue of collaborative data sharing among a number of parties to provide rich online services to their clients. We assume that each party hosts its data in a private cloud infrastructure but they collectively agree to certain well-defined set of accesses and access restrictions to one another's databases. We show how these can be used to derive access rules that become the basis of access control and query planning. In this paper we focus entirely on relational database sharing, and present efficient and near optimal heuristic algorithms for access rule derivation and query planning in spite of the NP-hardness and high complexity of the underlying problems.

Private IP addressing is commonly used in enterprise networks. Different enterprises or even different locations/business units of the same enterprise may use the same IP address ranges as long while those networks are separate. Consequently, during mergers and acquisitions, or network consolidations within an enterprise, overlapped and conflicted IP segments (subnets) arise frequently. These must be identified and resolved to allow communication between any pair of source and destination hosts in the merged networks. Furthermore, the combined network may unnecessarily use many disparate IP address ranges which increases the size of the routing tables and makes routing integrity verification difficult. In this paper, we identify different conflict scenarios and consider ways of resolving those conflicts to minimize manual changes and to minimize the routing table sizes. The problem turns out to be NP-hard and rather complex, and we devise effective heuristics to solve the problem. By taking some real-world examples, we show that by changing 6-8% of the subnet addresses the outlined method can effectively resolve the subnet conflicts. The scheme also reduces the number of subnet entries by 80-90% by consolidating the subnet entries, which significantly reduces the routing table sizes.

In this paper we develop a general framework for multi-sensor, heterogeneous sensing in collaborative wireless sensor networks (WSNs) that can be used in a variety of large scale monitoring applications. In order to achieve better tolerance against unstable wireless links and nodes with inadequate battery, it is important to consider distributed approaches for sampling rate adaptation. We show that the fully distributed mechanisms suffer from high convergence time, which make them difficult to implement in large-scale WSNs. To overcome this limitation, we next propose two alternate approaches. We perform extensive simulations to compare these schemes and argue their scalability and applicability in real world monitoring scenarios.

The current bulk transit systems (e.g., buses) and local perishable food distribution logistics, both suffer from significant fuel inefficiency along with food wastage due to quality degradation in the distribution pipeline. In this paper we present a mechanism that exploits automated electric vehicles (AEVs) in future smart cities and regions to provide both people transport and fresh food distribution that minimizes empty miles of the vehicles (and thus enhances transport efficiency) while meeting the constraints on passenger transit time and food freshness. We devise an optimization framework and show how it can be solved using genetic algorithms in order to handle dynamic demands for passenger transport/products, uncertain supply delays, and variations in product availability. Performance evaluations with extensive simulations show that flexibly deciding the AEV routes improves the transportation efficiency by ~24-78% whereas improves the delivery quality by ~2 times compared to the typical fixed routes/schedules used both by regular passenger bus services and by local distribution operations.

In this paper, we consider restricted data sharing between a set of parties that wish to provide some set of online services requiring such data sharing. Each party is assumed to store its data in private relational databases, and is given a set of mutually agreed set of authorization rules that specify access to attributes over individual relations or joins over relations owned by one or more parties. The access restrictions introduce significant additional complexity in rule enforcement and query planning as compared with a traditional distributed database environment. We examine the problem of minimum cost rule enforcement which simultaneously checks for the enforceability of each rule and generation of minimum cost plan of its execution. However, the paper is not focused on specific cost functions, but instead of efficient methods for enforcing rules in the face of access restrictions and inter-party data transfer needs. We propose an efficient heuristic algorithm for this minimal enforcement since the exact problem is NP-hard. In some cases, it is not possible to enforce the rules with the regular parties only. In such cases, we need help of trusted third parties (TPs). If all parties trust a single TP, such a party can enforce all unenforced rules, but it is desirable to use the TP minimally. We also consider the extended case where multiple TPs are required since not every regular party can trust a single TP.

In this paper we consider sensor networks for detecting contamination in urban water distribution systems. We assume that the sensor nodes are installed at connection points only (through the manholes) and are driven by super-capacitors charged by water flow. Although water systems may be affected by a large variety of contaminants, only a few sensors can be practically deployed. Thus many types of contaminants are sensed via “proxy sensing”, which may not be 100% reliable. In this paper we consider such a situation and examine the problem of collaborative adaptation of heterogeneous set of sensors in order to maximize contamination detection, especially during periods of almost zero natural water flow. The paper shows, through extensive simulations, that the proposed approach can drastically reduce the contamination reporting time from 31/2 hours to ~6 minutes, compared to the case without adaptation.

In this paper, we introduce a novel all-optical Data center networking (DCN) fabric, by leveraging the reconfigurability of the optical transceivers and switches, while dynamically changing the end-to-end optical routes to match the varying traffic demands. The dynamic flow scheduling along with their wavelength assignment turns out to be a NP-hard problem. We propose centralized heuristics for the inter-rack flow scheduling, by exploiting the optical wavelength division multiplexing, while minimizing the number of intermediate optical hops. Through extensive simulations, we show that the proposed architecture and flow scheduling reduces the network congestion by a factor of 15-18, compared to state-of-the-art part-time optical DCNs. For most of the traffic patterns, the proposed scheme can deliver >90% of the inter-rack traffic through direct optical communication.

In this paper, we introduce the concept of Water flow Driven Sensor Networks for leakage and contamination monitoring in urban water distribution systems. The unique aspect of our work is that the sensor network can be deployed in the underground water network with only access to connection points (through manholes) and driven only by water harvested energy so as to avoid access to AC power or need for frequent battery changes. The main problems addressed are (a) adaptation of the network to the available energy in order to maximize leak/contamination detection, and (b) minimal artificial water circulation or leakage to improve detectability during periods of almost zero natural water flow. The paper shows, through extensive simulations, that the proposed approach can drastically reduce the leakage/contamination reporting time (more than 3 hours to ~30 minutes), and the adaptation can reduce this circulation by ~33% and yet enhance the collected/transmitted data by 30%.

Handling of storage IO in modern operating systems assumes that such devices are slow and CPU cycles are valuable. Consequently, to effectively exploit the underlying hardware resources, for example, CPU cycles, storage bandwidth and the like, whenever an IO request is issued to such device, the requesting thread is switched out in favor of another thread that may be ready to execute. Recent advances in nonvolatile storage technologies and multicore CPUs make both of these assumptions increasingly questionable, and an unconditional context switch is no longer desirable. In this article, we propose a novel mechanism called SmartCon, which intelligently decides whether to service a given IO request in interrupt-driven manner or busy-wait--based manner based on not only the device characteristics but also dynamic parameters such as IO latency, CPU utilization, and IO size. We develop an analytic performance model to project the performance of SmartCon for forthcoming devices. We implement SmartCon mechanism on Linux 2.6 and perform detailed evaluation using three different IO devices: Ramdisk, low-end SSD, and high-end SSD. We find that SmartCon yields up to a 39% performance gain over the mainstream block device approach for Ramdisk, and up to a 45% gain for PCIe-based SSD and SATA-based SSDs. We examine the detailed behavior of TLB, L1, L2 cache and show that SmartCon achieves significant improvement in all cache misbehaviors.

The 7th ACM International Workshop on Cyber Security Analytics, Intelligence and Automation (SafeConfig) is held as part of ACM Computer and Communications Security CCS 2014. SafeConfig 14, following many successful preceding workshops, is concerned with developing new security techniques and approaches that offer proactive, intelligent and a holistic security analytics based on analyzing the system artifacts including system traces, configurations, logs, incident reports, alarms and network traffic. Scalable analytics techniques are essential to handle large volumes of data and to normalize, model, integrate, analyze and respond to threats in real time.

In this paper, we propose a novel architecture called Energy Aware Disaster Recovery Network using WiFi Tethering (E-DARWIN). The underlying idea is to make use of WiFi Tethering technology ubiquitously available on wireless devices, like smartphones and tablets, to set up an ad hoc network for data collection in disaster scenarios. To this end, we design novel mechanisms, which aid in autonomous creation of the ad hoc network, distribution of data capturing task among the devices, and collection of data with minimum delay. Specifically, we design and implement a distributed coalition formation game for distributing the data capturing task among wireless devices based on their capabilities, available energy, and network participation for higher network lifetime. Finally, we evaluate the performance of the proposed architecture using a prototype application implemented on Android platform and large-scale simulations.

In this paper we propose a flexible and scalable distributed storage framework called flex Store that can adapt to variations in available or consumable power and demonstrate its performance in the context of reduplicated virtual machine disks. We propose and investigate smart control techniques in order to cope with the power constraints either introduced as a result of increasing node density in the storage arrays (consumable power constraints) or introduced when a mix of renewable (green) and conventional (brown) energy sources are used to power the data enter. The key component in the proposed storage framework is the policy engine which is a software layer that provides interfaces to define performance requirements of the applications (and also energy related policies). The policy engine enforces those policies in the storage system by adjusting the allocation of storage resources. The experimental results demonstrate the ability of the framework to dynamically adapt to the changes in workload and power constraints and minimize performance impacts. Our evaluation of the prototype shows that the adaptive replication mechanisms can reduce the IO latencies by around 65% during energy plenty situations and the impact of adaptation actions on IO latencies during energy constrained situations is reduced by more than 40% compared to the case without the adaptive replication and optimized adaptation mechanisms.

The Domain Name System (DNS) provides a critical service for the Internet -- mapping of user-friendly domain names to their respective IP addresses. Yet, there is no standard set of metrics quantifying the Quality of Domain Name Service (QoDNS), let alone a thorough evaluation of it. This article attempts to fill this gap from the perspective of a DNS proxy/cache, which is the bridge between clients and authoritative servers. We present an analytical model of DNS proxy operations that offers insights into the design trade-offs of DNS infrastructure and the selection of critical DNS parameters. Due to the critical role DNS proxies play in QoDNS, they are the focus of attacks including cache poisoning attack. We extend the analytical model to study DNS cache poisoning attacks and their impact on QoDNS metrics. This analytical study prompts us to present Domain Name Cross-Referencing (DoX), a peer-to-peer systems for DNS proxies to cooperatively defend cache poisoning attacks. Based on QoDNS, we compare DoX with the cryptography-based DNS Security Extension (DNSSEC) to understand their relative merits.

In this paper we consider the situation where a set of enterprises need to collaborate to provide rich services to their clients. Such collaboration often requires controlled access to each other's data, which we assume is stored in standard relational form. The access control is provided by a set o

Secure socket layer (SSL) is the most popular protocol used in the Internet for facilitating secure communications. In this retrospective, we summarize our original paper which analyzed the performance and architectural impact of SSL on the servers and provided insights into the functioning and acceleration of SSL. In addition, we describe advancements in the area that have occurred on this topic and also discuss future research opportunities.

Secure socket layer (SSL) is the most popular protocol used in the Internet for facilitating secure communications. In this paper, we analyze the performance and architectural impact of SSL on the servers in terms of various parameters such as throughput, utilization, cache sizes, cache miss ratios, number of processors, control dependencies, file access sizes, bus transactions, network load, etc. The major conclusions from this study are as follows: The use of SSL increases computational cost of the transactions by a factor of 5-7. SSL transactions do not benefit much from a larger L2 cache, but a larger LI cache would be helpful. A complex logic for handling control dependencies is not useful for SSL transaction as the frequency of branches is very low. Because SSL workload is highly CPU bound, it may be possible to enhance SSL performance by using a number of other architectural features as well.

The sustainability concerns of Information Technology (IT) go well beyond energy-efficient computing and require techniques for minimizing environmental impact of IT infrastructure over its entire life-cycle. Traditionally, IT infrastructure is overdesigned at all levels from chips to entire data centers and ecosystem; the paradigm explored in this article is to replace overdesign with rightsizing coupled with smarter control, henceforth referred to as Energy-Adaptive Computing or EAC. The article lays out the challenges of EAC in various environments in terms of the adaptation of the workload and the infrastructure to cope with energy and cooling deficiencies. The article then focuses on implementing EAC in a data center environment, and addresses the problem of simultaneous energy demand and energy supply regulation at multiple levels, work, from servers to the entire data center. The proposed control scheme adapts the assignments of tasks to servers in a way that can cope with the varying energy limitations. The article also presents some experimental results to show how the scheme can continue to meet Quality of Service (QoS) requirements of tasks under energy limitations.

The Domain Name System (DNS) is a foundational component of today's Internet for mapping Internet names to addresses. With the DNS Security Extensions (DNSSEC) DNS responses can be cryptographically verified to prevent malicious tampering. The protocol complexity and administrative overhead associated with DNSSEC can significantly impact the potential for name resolution failure. We present metrics for assessing the quality of a DNSSEC deployment, based on its potential for resolution failure in the presence of DNSSEC misconfiguration. We introduce a metric to analyze the administrative complexity of a DNS configuration, which contributes to its failure potential. We then discuss a technique which uses soft anchoring to increase robustness in spite of misconfigurations. We analyze a representative set of production signed DNS zones and determine that 28% of the validation failures we encountered would be mitigated by the soft anchoring technique we propose.

The increasing energy demand coupled with emerging sustainability concerns requires a re-examination of power/thermal issues in data centers from the perspective of short term energy deficiencies. Such energy deficient scenarios arise for a variety of reasons including variable energy supply from renewable sources and inadequate power, thermal and cooling capacities. In this paper we propose a hierarchical control scheme to adapt assignments of tasks to servers in a way that can cope with the varying energy limitations and still provide necessary QoS. The rescheduling of tasks on different servers has direct (migration related) and indirect (changed traffic patterns) network energy impacts that we also consider. We show the stability of our scheme and evaluate its performance via detailed simulations and experiments.

No abstract available.

This paper introduces a closed-loop control algorithm to coordinate power management of memory ranks and thereby achieve power savings beyond independent rank power management while bounding the throughput degradation.

Fueled by burgeoning online services, power and thermal issues are becoming crucial both in terms of utility costs and environmental impact. In this paper, we motivate an approach that puts power/thermal issues at the heart of distributed computing, and strives to alter the behavior of the computation in order to dynamically adjust to changing energy constraints. The paper discusses three distributed energy adaptation paradigms and shows why they are essential for sustainable computing. The paper also compares and contrasts two basic techniques for energy adaptation in servers, namely power capping and use of sleep modes.

The increasing power consumption of computing infrastructure and power density in computing devices is becoming a major concern from a variety of perspectives ranging from Moore's Law scaling to sustainability. Addressing these concerns results in designs that are much leaner in terms of their energy, power and thermal capabilities, and thus require much smarter control to stay within the provided budgets. This leads to the notion of energy adaptive computing where the power management needs to go well beyond energy efficiency and handle situations of insufficient energy via a variety of adaptation techniques. In this paper we address the problem of simultaneous power demand and supply regulation for multilevel systems such as data centers. We expose challenges in such controls and examine the performance of a simple control scheme designed to provide a stable control via demand migrations.

The domain name system (DNS) is critical to Internet functionality. The availability of a domain name refers to its ability to be resolved correctly. We develop a model for server dependencies that is used as a basis for measuring availability. We introduce the minimum number of servers queried (MSQ) and redundancy as availability metrics and show how common DNS misconfigurations impact the availability of domain names. We apply the availability model to domain names from production DNS and observe that 6.7% of names exhibit sub-optimal MSQ, and 14% experience false redundancy. The MSQ and redundancy values can be optimized by proper maintenance of delegation records for zones.

The domain name system (DNS) is integral to today's Internet. Name resolution for a domain is often dependent on servers well outside the control of the domain's owner. In this paper we propose a formal model for analyzing the name dependencies inherent in DNS, based on protocol specification and actual implementations. We derive metrics to quantify the extent to which domain names affect other domain names. It is found that under certain conditions, the name resolution for over one-half of the queries exhibits influence of domains not expressly configured by administrators. This result serves to quantify the degree of vulnerability of DNS due to dependencies that administrators are unaware of. The model presented in the paper also shows that the set of domains whose resolution affects a given domain name is much smaller than previously thought. The model also shows that with caching of NS target addresses, the number of influential domains expands greatly, thereby making the DNS infrastructure more vulnerable.

The border gateway protocol (BGP) is known to take a long time to converge to a steady state following the failure of BGP routers or inter-router links. This has resulted in extensive analysis of BGP convergence delay and a number of schemes have been proposed to reduce this delay. But the convergence delay is a network centric metric and the end-user relevant effects of a failure are better characterized by the packet losses. In this paper we study BGP convergence from the packet delivery perspective and show that a reduction in the convergence delay does not necessarily translate into an improvement in packet delivery. Our measurements provide insights into which BGP modifications are likely to decrease packet loss, and how any shortcomings can be rectified. We also modify a couple of existing techniques, and show how to reduce the packet losses.

The domain name system (DNS) provides a critical service for the Internet -mapping of user-friendly domain names to their respective IP addresses. Yet, there is no standard set of metrics quantifying the quality of domain name service or QoDNS, let alone a thorough evaluation of it. This paper attempts to fill this gap from the perspective of a DNS proxy/cache, which is the bridge between clients and authoritative servers. We present an analytical model of DNS proxy operations that offers insights into the design tradeoffs of DNS infrastructure and the selection of critical DNS parameters. After validating our model against simulation results, we extend it to study the impact of DNS cache poisoning attacks and evaluate various DNS proposals with respect to the QoDNS metrics. In particular, we compare the performance of two newly proposed DNS security solutions: one based on cryptography and one using collaborative overlays.

Border gateway protocol (BGP) is the standard routing protocol used in the Internet for routing packets between the autonomous systems (ASes). It is known that BGP can take hundreds of seconds to converge after isolated failures. We have also observed that the convergence delay can be even greater for large-scale failures. In this study, we first investigate some of the factors affecting the convergence delay and their relative impacts. We observe that the minimum route advertisement interval (MRAI) and the processing overhead at the routers during the re-convergence have a significant effect on the BGP recovery time. We propose a couple of new schemes to reduce processing overload at BGP routers during large failures, which in turn leads to decreased convergence delays. We show that these schemes combined with the tuning of the MRAI value decrease the BGP convergence delay significantly, and can thus limit the impact of large scale failures in the Internet

Border gateway protocol (BGP) is the standard routing protocol between various autonomous systems (AS) in the Internet. In the event of a failure, BGP may repeatedly withdraw some routes and advertise new ones until a stable state is reached. It is known that the corresponding recovery time could stretch into hundreds of seconds or more for isolated Internet outages and lead to high packet drop rates. In this paper we characterize BGP recovery time under large-scale failure scenarios, perhaps those caused by disastrous natural or man-made events. We show that the recovery time depends on a variety of topological parameters and can be substantial for massive failures. The study provides guidelines on reducing the impact of BGP convergence delay on the Internet.

The mapping service provided by the Domain Name System (DNS) is fundamental not only to the health of the Internet but also to the protection and integrity of the data. Recently, the DNS infrastructure has suffered several malicious attacks including DNS cache poisoning, which causes the DNS to return false name-to-IP mappings and can be used as a foothold for more insidious attacks. This paper proposes DoX, a peer-to-peer based scheme, to detect and correct inaccurate DNS records caused by cache poisoning attacks. DoX also helps DNS servers to improve cache consistency by detecting and removing obsolete records. DoX does not require modifications to the current infrastructure and can be deployed quickly. It does not use cryptographic techniques and thus does not suffer from the key management and processing overhead issues of those techniques.

The border gateway protocol (BGP) has been known to suffer from large convergence delays after failures. We have also found that the impact of a failure rises sharply with the size of the failure. In this paper we present and evaluate a speculative route invalidation scheme aimed at reducing the convergence delays for large-scale failures. Our scheme collects statistics from BGP updates received at a router and identifies Autonomous Systems (ASes) that are likely to be "unstable". Routes that contain these ASes are marked as invalid and not propagated further. This cuts down the number of invalid routes during the convergence process and results in a significant improvement in the convergence delay.

In this paper, we study the performance of a clustered DBMS running an on-line transaction processing (OLTP) workload and using TCP/IP over Ethernet as a "unified fabric" for inter-process communication. The study is based on a comprehensive simulation model of such systems, called DCLUE. It is found that while the protocol overhead has a large impact on performance, the end to end latency has significantly lesser impact. The results suggest that very low-latency protocol offload may not be very beneficial for clustered OLTP workloads, but proper QoS treatment of messages is important.

This paper presents a detailed modeling of TCP offload in the context of SPECweb99 web-server benchmark in order to assess the impact of offload on the overall server performance. A number of possibilities with respect to offload are considered, including the offload location, data/connection offload and direct/iSCSI based storage. The results indicate that a carefully designed offload engine coupled with a lean host interface can more than double the SPECweb99 performance.

In this paper we introduce a non-uniform random-graph model for studying the evolution of ad hoc peer-to-peer (P2P) communities. The model provides control over the nodal degree distribution. We study basic properties such as reachability from a given node and overall traffic handled by a node using an analytical approach. The analytic model can be used in conjunction with a simulation model to study detailed performance tradeoffs in a P2P file-sharing network.

As microprocessors scale rapidly in frequency, the design of fast and efficient interconnects becomes extremely important for low latency data access and high performance. We evaluate a technique for reducing the interconnect width by exploiting the spatial and temporal locality in communication transfers (addresses & data). The width reduction implies a number of other advantages including higher operating frequency, reduced pin-count, lower chip & board cost, etc. We evaluate the effectiveness of the proposed scheme by performing trace-driven simulations for two well-known commercial server workloads (SPECWeb99 and TPC-C). We also study the sensitivity of the compression hit ratio with respect to the number of bits compressed, size of the encoding/decoding table used and the replacement policy. The results indicate that the proposed technique has a potential to reduce address bus width in most cases and data bus widths in some cases while maintaining equal or better performance than in the uncompressed case.

Popularized by Napster and Gnutella file sharing solutions, peer-to-peer (P2P) computing has suddenly emerged at the forefront of Internet computing. The basic notion of cooperative computing and resource sharing has been around for quite some time, although these new applications have opened up possibilities of very flexible web-based information sharing. This article provides a frame-work for classifying current and future P2P technologies. The main motivation for the classification is to identify basic characteristics of P2P applications so that the infrastructure to support P2P computing can concentrate on these basic characteristics.

We provide a generic model to characterize transactional behavior in an e-commerce environment which can be applied to business-to-business (B2B) e-commerce sites, Business-to-consumer (B2C) e-commerce sites, and other (e.g., simple browsing) type of Web-sites. The ideas presented in this paper have been derived based upon detailed examination of the e-commerce server architecture and traffic from a couple of sites. The paper also studies the characterization of embedded requests. We also discuss how the transactional and embedded request models can be used in generation of artificial front-end e-commerce traffic in the laboratory for resource engineering of servers.

The performance of servers for network-intensive workloads such as web services and online transaction processing applications depends on the effective utilization of the processor caches. A detailed analysis of the cache space utilization of web workloads shows us that several memory addresses are referenced only once during their lifetime in the cache. These references frequently reside in the cache for a long time contributing to the pollution of cache. The most commonly adopted least-recently-used (LRU) replacement scheme does not exploit this characteristic. In this paper, we propose an alternative block replacement policy called Single-Touch Aware Replacement (STAR) algorithm. This algorithm predicts blocks that will potentially be referenced only once and replaces them early enough to improve cache efficiency. The STAR scheme was implemented in a trace-driven cache simulator and the performance with several commercial workloads was analyzed. The use of the STAR algorithm results in up to 20% improvement in cache performance for web workloads (SPECweb96, SPECweb99) and up to 5% improvement in online transaction processing (TPC-C) workloads.

This paper describes Geist, a traffic generator for stress testing of web servers. The generator provides a large number of dialable parameters that allow traffic characteristics to range from simple static web-page browsing to the transactional traffic seen by e-commerce front end servers. Unlike other traffic generators, our generator concentrates on the characteristics of the aggregate traffic arriving at the server, which allows for better control of the scaling properties of the traffic and a more scalable generation. In this paper, we describe the traffic characterization and generation process that Geist is based on. We also present the performance of our current implementation, instances of its usage and directions for future work.

Secure socket layer (SSL) is the most popular protocol used in the Internet for facilitating secure communications. In this paper, we analyze the performance and architectural impact of SSL on the servers in terms of various parameters such as throughput, utilization, cache sizes, cache miss ratios, number of processors, control dependencies, file access sizes, bus transactions, network load, etc. The major conclusions from this study are as follows: The use of SSL increases computational cost of the transactions by a factor of 5-7. SSL transactions do not benefit much from a larger L2 cache, but a larger L1 cache would be helpful. A complex logic for handling control dependencies is not useful for SSL transaction as the frequency of branches is very low. Because SSL workload is highly CPU bound, it may be possible to enhance SSL performance by using a number of other architectural features as well.

The paper describes a queuing network model for a multiprocessor system running a static Web workload such as SPECweb96. The model includes architectural details of the Web server in terms of multilevel cache hierarchy, processor bus, memory pipeline, PCI bus based I/O subsystem, and bypass I/O-memory path for DMA transfers. The model is based on detailed measurements from a baseline system and a few of its variants. The model operates at the Web transaction level, and does not explicitly model the CPU core or the caching hierarchy. Yet, the model predicts the performance impact of low level features such as number of processors, processor speeds, cache sizes and latencies, memory latencies, higher level caches, sector prefetching, etc. The model shows an excellent match with measured results. Because of many features that are difficult to handle analytically, the default solution technique is simulation. However, the paper also proposes a simple hybrid approach that can significantly speed up the solution without affecting the accuracy appreciably. The model has also been extended to handle clusters of symmetric multiprocessor systems with both centralized and distributed memories.

The goal of the paper is to provide a methodology for determining bandwidth requirements for various hardware components of a World Wide Web server. The paper assumes a traditional symmetric multiprocessor (SMP) architecture for the Web server, although the same analysis applies to an SMP node in a cluster. The paper derives formulae for bandwidth demands for memory, processor data bus, network adapters, disk adapters, I/O memory paths, and I/O buses. Since the Web workload characteristics vary widely, three sample workloads are considered for illustrative purposes: 1) standard SPECweb96; 2) a SPECweb96-like workload that assumes dynamic data and retransmissions; and 3) WebProxy, which models a Web proxy server that does not do much caching, and thus has rather severe requirements. The results point to a few general conclusions regarding Web workloads. In particular, reduction in memory/data bus bandwidth by using the virtual interface architecture (VIA) is very desirable, and the connectivity needs may go well beyond the capabilities of traditional systems based on the traditional PCI-bus. Web workloads also demand a significantly higher memory bandwidth than data bus bandwidth and this disparity is expected to increase with the use of VIA. Also, the current efforts to offload TCP/IP processing may require a larger headroom in I/O subsystem bandwidth than in the processor-memory subsystem.

Motivated by field data which showed a large number of link changeovers and incidences of link oscillations between in-service and out-of-service states in common channel signalling (CCS) networks, a number of analyses of the link error monitoring procedures in the SS7 protocol were performed by the authors. This paper summarizes the results obtained thus far and include the following: (a) results of an exact analysis of the performance of the error monitoring procedures under both random and bursty errors; (b) a demonstration that there exists a range of error rates within which the error monitoring procedures of SS7 may induce frequent changeovers and changebacks; (c) an analysis of the performance of the SS7 level-2 transmission protocol to determine the tolerable error rates within which the delay requirements can be met; (d) a demonstration that the tolerable error rate depends strongly on various link and traffic characteristics, thereby implying that a single set of error monitor parameters will not work well in all situations; and (e) some recommendations on a customizable/adaptable scheme of error monitoring with a discussion on their implementability. These issues may be particularly relevant in the presence of anticipated increases in SS7 traffic due to widespread deployment of advanced intelligent network (AIN) and personal communications service (PCS) as well as for developing procedures for high-speed SS7 links currently under consideration by standards bodies.>

Analyzing the performance of real-time control systems featuring mechanisms for online recovery from software faults is discussed. The application is assumed to consist of a number of interacting cyclic processes. The underlying hardware is assumed to be a multiprocessor, possibly with a separate control processor. The software structure is assumed to use design diversity along with forward and/or backward recovery. A detailed but efficiently solvable model for predicting various performance and reliability characteristics is developed. One of the key ideas used in modeling is hierarchical decomposition, which enables computation of level-oriented performance parameters in an efficient manner. The model is general, and adaptable for a number of useful special cases.>

A formal approach is presented for the analysis and synthesis of robust data structures. The entire data structure is viewed as a collection of data elements related via some attributes. The relationships are specified by a set of axioms in first-order logic. Faults in attributes invalidate some of the axioms. The invalidated axioms are used to detect and correct the faulty attributes. The authors show how detection and correction can be localized to small portions of the data structure, thereby allowing concurrent repair in several disjoint portions. This property makes local correction attractive for B-trees and other structures used in databases. They then show how the ideas developed for attaining structural integrity can be applied to achieve data integrity as well.>

In this paper, we consider the application level performance modeling of parallel machines consisting of a large number of processing elements (PE's) connected in some regular structure such as mesh, tree, hypercube, etc. There are K problem types, each arriving according to a Poisson process, and each of which needs a PE substructure of some given size and topology. Thus several problems can run on the machine simultaneously. It is desired to characterize the performance of such a system under various types of allocation schemes.

Backward error recovery is a popular technique for recovery from unexpected system failures. In a concurrent processing environment, the synchronisation constraints and the propagation of erroneous information between processes tend to make recovery very complex and expensive. In this paper we present a detailed analysis of error propagation and introduce a new classification for variables based upon their error-propagation characteristics. Necessary and sufficient conditions to ensure a finite upper bound on the computation discarded in the course of error recovery are developed. Several other recovery-related issues are also discussed.

The paper examines the problem of counting all interferences between a set of rectangular paths. Tight upper bounds on the number of overlaps and intersections between two or more rectangular paths are derived both for the general case and the special cases where the backtracking is not allowed in one or both directions. An optimal algorithm for counting all interferences for both the special and general cases is presented.