Swedish Institute of Computer Science

Currently, most of the existing data centers use chilled air to remove the heat produced by the servers. However, liquids have generally better heat dissipation capabilities than air, thus liquid cooling systems are expected to become a standard choice in future data centers. Designing and managing these cooling units benefit from having control-oriented models that can accurately describe the thermal status of both the coolant and the heat sources. This manuscript derives a control-oriented model of liquid immersion cooling systems, i.e., systems where servers are immersed in a dielectric fluid having good heat transfer properties. More specifically, we derive a general lumped-parameters gray box dynamical model that mimics energy and mass transfer phenomena that occur between the main components of the system. The proposed model is validated against experimental data gathered during the operation of a proof-of-concept immersion cooling unit, showing good approximation capabilities.

The IETF IPv6 over the TSCH mode of IEEE802.15.4e (6TiSCH) working group has standardized a set of protocols to enable low power industrial-grade IPv6 networks. 6TiSCH proposes a protocol stack rooted in the Time Slotted Channel Hopping (TSCH) mode of the IEEE802.15.4-2015 standard, supports multi-hop topologies with the IPv6 Routing Protocol for Low-Power and Lossy Networks (RPL) routing protocol, and is IPv6-ready through 6LoWPAN. 6TiSCH has defined the missing control plane protocols to match link-layer resources to the routing topology and application communication needs. 6TiSCH has also defined a secure light-weight join processes combining link-layer security features (through Counter with CBC-MAC (CCM*)) with a secure joining procedure using the Constrained Application Protocol (CoAP). This tutorial provides a comprehensive overview of the 6TiSCH architecture and protocol suite, including the 6TiSCH Operation Sublayer (6top), the 6top Protocol (6P), and how it uses 6LoWPAN, IP-in-IP encapsulation, and RPL. This document is meant to be used both as a primer, and as a reference. It is tailored to the advanced researcher and engineer implementing and building upon IETF 6TiSCH specifications.

In the era of big data, considerable research focus is being put on designing efficient algorithms capable of learning and extracting high-level knowledge from ubiquitous data streams in an online fashion. While, most existing algorithms assume that data samples are drawn from a stationary distribution, several complex environments deal with data streams that are subject to change over time. Taking this aspect into consideration is an important step towards building truly aware and intelligent systems. In this paper, we propose GNG-A, an adaptive method for incremental unsupervised learning from evolving data streams experiencing various types of change. The proposed method maintains a continuously updated network (graph) of neurons by extending the Growing Neural Gas algorithm with three complementary mechanisms, allowing it to closely track both gradual and sudden changes in the data distribution. First, an adaptation mechanism handles local changes where the distribution is only non-stationary in some regions of the feature space. Second, an adaptive forgetting mechanism identifies and removes neurons that become irrelevant due to the evolving nature of the stream. Finally, a probabilistic evolution mechanism creates new neurons when there is a need to represent data in new regions of the feature space. The proposed method is demonstrated for anomaly and novelty detection in non-stationary environments. Results show that the method handles different data distributions and efficiently reacts to various types of change.

Society is becoming increasingly dependent on IT services. One example is the dependence of retailers on electronic payment services. This article investigates the terms and conditions offered by three electronic payment service providers, finding that they only guarantee best effort availability. As potential mitigation, five cyber insurance policies are studied from the perspective of coverage of electronic payment service outages. It is concluded that cyber insurance does indeed give some protection, but that coverage differs between insurers and between different policy options offered. Thus, a retailer who wishes to purchase cyber insurance should take care to understand what is on offer and actively select appropriate coverage.

With the growth of the Mobile Internet, people have become active in both the online and offline worlds. Investigating the relationships between users’ online and offline behaviors is critical for personalization and content caching, as well as improving urban planning. Although some studies have measured the spatial properties of online social relationships, there have been few in-depth investigations of the relationships between users’ online content browsing behaviors and their real-life locations. This paper provides the first insight into the geospatial properties of online content browsing behaviors from the perspectives of both geographical regions and individual users. We first analyze the online browsing patterns across geographical regions. Then, a multilayer-network-based model is presented to discover how inter-user distances affect the distributions of users with similar online browsing interests. Drawing upon results from a comprehensive study of users of three popular online content services in a metropolitan city in China, we achieve a broad understanding of the general and specific geospatial properties of users’ various preferences. Specifically, users with similar online browsing interests exhibit, to a large extent, strong geographic correlations, and different services exhibit distinct geospatial properties in terms of their usage patterns. The results of this work can potentially be exploited to improve a vast number of applications.

Certificate-based Public Key Infrastructure (PKI) schemes are used to authenticate the identity of distinct nodes on the Internet. Using certificates for the Internet of Things (IoT) can allow many privacy sensitive applications to be trusted over the larger Internet architecture. However, since IoT devices are typically resource limited, full sized PKI certificates are not suitable for use in the IoT domain. This work outlines our approach in compressing standards-compliant X.509 certificates so that their sizes are reduced and can be effectively used on IoT nodes. Our scheme combines the use of Concise Binary Object Representation (CBOR) and also a scheme that compresses all data that can be implicitly inferenced within the IoT sub-network. Our scheme shows a certificate compression rate of up to ~30%, which allows effective energy reduction when using X.509-based certificates on IoT platforms.

Enterprise architecture (EA) has been established as a discipline to cope with the complex interactions of business operations and technology. Models, i.e., formal descriptions in terms of diagrams and views, are at the heart of the approach. Though it is widely thought that such architecture models can contribute to improved understanding and decision making, this proposition has not rigorously been tested. This article describes an experiment conducted with a real EA model and corresponding real traditional documents, investigating whether the model or the documents lead to better and faster understanding. Understanding is interesting to study, as it is a prerequisite to other EA uses. The subjects (N=98) were officer cadets, and the experiment was carried out using a comprehensive description of military Close Air Support capability either (1) in the form of a MODAF model or (2) in the form of traditional documents. Based on the results, the model seems to lead to better, though not faster, understanding.

The Internet of Things (IoT) consists of resource-constrained devices (e.g., sensors and actuators) which form low power and lossy networks to connect to the Internet. With billions of devices deployed in various environments, IoT is one of the main building blocks of future Internet of Services (IoS). Limited power, processing, storage and radio dictate extremely efficient usage of these resources to achieve high reliability and availability in IoS. Denial of Service (DoS) and Distributed DoS (DDoS) attacks aim to misuse the resources and cause interruptions, delays, losses and degrade the offered services in IoT. DoS attacks are clearly threats for availability and reliability of IoT, and thus of IoS. For highly reliable and available IoS, such attacks have to be prevented, detected or mitigated autonomously. In this study, we propose a comprehensive investigation of Internet of Things security for reliable Internet of Services. We review the characteristics of IoT environments, cryptography-based security mechanisms and D/DoS attacks targeting IoT networks. In addition to these, we extensively analyze the intrusion detection and mitigation mechanisms proposed for IoT and evaluate them from various points of view. Lastly, we consider and discuss the open issues yet to be researched for more reliable and available IoT and IoS.

Managing and balancing load in distributed systems remains a challenging problem in resource management, especially in networked systems where scalability concerns favour distributed and dynamic approaches. Distributed methods can also integrate well with centralised control paradigms if they provide high‐level usage statistics and control interfaces for supporting and deploying centralised policy decisions. We present a general method to compute target values for an arbitrary metric on the local system state and show that autonomous rebalancing actions based on the target values can be used to reliably and robustly improve the balance for metrics based on probabilistic risk estimates. To balance the trade‐off between balancing efficiency and cost, we introduce 2 methods of deriving rebalancing actuations from the computed targets that depend on parameters that directly affects the trade‐off. This enables policy level control of the distributed mechanism based on collected metric statistics from network elements. Evaluation results based on cellular radio access network simulations indicate that load balancing based on probabilistic overload risk metrics provides more robust balancing solutions with fewer handovers compared to a baseline setting based on average load.

Today, with rapidly developing technology and changing business models, organizations face rapid changes in both internal and external environments. To be able to rapidly respond to such changing environments, integration of software systems has become a top priority for many organizations. However, despite extensive use of software systems integration, quantitative methods for estimating the business value of such integrations are still missing. Using Data Envelopment Analysis (DEA) and the microeconomic concept of marginal rates, this study proposes a method for quantifying the effects of enterprise integration on the firm performance. In the paper, we explain how DEA can be used to evaluate the marginal benefits of enterprise integration. Our proposed method is to measure and compare the productive efficiency of firms using enterprise integration, specifically by relating the benefits produced to the resources consumed in the process. The method is illustrated on data collected from 12 organizations. The defined method has a solid theoretical foundation, eliminating the need for a priori information about the relationship between different measures. Furthermore, the framework could be used not only to quantify the business value of enterprise integration, but also to estimate trade-offs and impacts of other subjective managerial goals on the results. The major limitation of the proposed method is the absence of a comprehensive theory relating IT architecture changes to organizational outcomes. The underlying model is strongly dependent on the relevancy and accuracy of the included variables, as well as number of data units, introducing uncertainties to the outcomes of the model.

A wide gap exists between the state of the art in developing Wireless Sensor Network (WSN) software and current practices concerning the design, execution, and maintenance of business processes. WSN software is most often developed based on low-level OS abstractions, whereas business process development leverages high-level languages and tools. This state of affairs places WSNs at the fringe of industry. The makeSense system addresses this problem by simplifying the integration of WSNs into business processes. Developers use BPMN models extended with WSN-specific constructs to specify the application behavior across both traditional business process execution environments and the WSN itself, which is to be equipped with application-specific software. We compile these models into a high-level intermediate language—also directly usable by WSN developers—and then into OS-specific deployment-ready binaries. Key to this process is the notion of meta-abstraction, which we define to capture fundamental patterns of interaction with and within the WSN. The concrete realization of meta-abstractions is application-specific; developers tailor the system configuration by selecting concrete abstractions out of the existing codebase or by providing their own. Our evaluation of makeSense shows that i) users perceive our approach as a significant advance over the state of the art, providing evidence of the increased developer productivity when using makeSense; ii) in large-scale simulations, our prototype exhibits an acceptable system overhead and good scaling properties, demonstrating the general applicability of makeSense; and, iii) our prototype—including the complete tool-chain and underlying system support—sustains a real-world deployment where estimates by domain specialists indicate the potential for drastic reductions in the total cost of ownership compared to wired and conventional WSN-based solutions. IEEE

The automatic detection and classification of stance (e.g., certainty or agreement) in text data using natural language processing and machine-learning methods creates an opportunity to gain insight into the speakers’ attitudes toward their own and other people’s utterances. However, identifying stance in text presents many challenges related to training data collection and classifier training. To facilitate the entire process of training a stance classifier, we propose a visual analytics approach, called ALVA, for text data annotation and visualization. ALVA’s interplay with the stance classifier follows an active learning strategy to select suitable candidate utterances for manual annotaion. Our approach supports annotation process management and provides the annotators with a clean user interface for labeling utterances with multiple stance categories. ALVA also contains a visualization method to help analysts of the annotation and training process gain a better understanding of the categories used by the annotators. The visualization uses a novel visual representation, called CatCombos, which groups individual annotation items by the combination of stance categories. Additionally, our system makes a visualization of a vector space model available that is itself based on utterances. ALVA is already being used by our domain experts in linguistics and computational linguistics to improve the understanding of stance phenomena and to build a stance classifier for applications such as social media monitoring.

We estimate end-to-end service metrics from network device statistics. Our approach is based upon statistical, supervised learning, whereby the mapping from device-level to service-level metrics is learned from observations, i.e., through monitoring the system. The approach enables end-to-end performance prediction without requiring an explicit model of the system, which is different from traditional engineering techniques that use stochastic modeling and simulation. The fact that end-to-end service metrics can be estimated from local network statistics with good accuracy in the scenarios we consider suggests that service-level properties are “encoded” in network-level statistics. We show that the set of network statistics needed for estimation can be reduced to a set of measurements along the network path between client and service backend, with little loss in estimation accuracy. The reported work is largely experimental and its results have been obtained through testbed measurements from a video streaming service and a KV store over an OpenFlow network .

Selecting sourcing options for software assets and components is an important process that helps companies to gain and keep their competitive advantage. The sourcing options include: in-house, COTS, open source and outsourcing. The objective of this paper is to further refine, extend and validate a solution presented in our previous work. The refinement includes a set of decision-making activities, which are described in the form of a process-line that can be used by decision-makers to build their specific decision-making process. We conducted five case studies in three companies to validate the coverage of the set of decision-making activities. The solution in our previous work was validated in two cases in the first two companies. In the validation, it was observed that no activity in the proposed set was perceived to be missing, although not all activities were conducted and the activities that were conducted were not executed in a specific order. Therefore, the refinement of the solution into a process-line approach increases the flexibility and hence it is better in capturing the differences in the decision-making processes observed in the case studies. The applicability of the process-line was then validated in three case studies in a third company.

We predict performance metrics of cloud services using statistical learning, whereby the behaviour of a system is learned from observations. Specifically, we collect device and network statistics from a cloud testbed and apply regression methods to predict, in real-time, client-side service metrics for video streaming and key-value store services. Results from intensive evaluation on our testbed indicate that our method accurately predicts service metrics in real time (mean absolute error below 16% for video frame rate and read latency, for instance). Further, our method is service agnostic in the sense that it takes as input operating systems and network statistics instead of service-specific metrics. We show that feature set reduction significantly improves the prediction accuracy in our case, while simultaneously reducing model computation time. We find that the prediction accuracy decreases when, instead of a single service, both services run on the same testbed simultaneously or when the network quality on the path between the server cluster and the client deteriorates. Finally, we discuss the design and implementation of a real-time analytics engine, which processes streams of device statistics and service metrics from testbed sensors and produces model predictions through online learning.

The pay-as-you-go pricing model and the illusion of unlimited resources in the Cloud initiate the idea to provision services elastically. Elastic provisioning of services allocates/de-allocates resources dynamically in response to the changes of the workload. It minimizes the service provisioning cost while maintaining the desired service level objectives (SLOs). Model-predictive control is often used in building such elasticity controllers that dynamically provision resources. However, they need to be trained, either online or offline, before making accurate scaling decisions. The training process involves tedious and significant amount of work as well as some expertise, especially when the model has many dimensions and the training granularity is fine, which is proved to be essential in order to build an accurate elasticity controller. In this paper, we present OnlineElastMan, which is a self-trained proactive elasticity manager for cloud-based storage services. It automatically evolves itself while serving the workload. Experiments using OnlineElastMan with Cassandra indicate that OnlineElastMan continuously improves its provision accuracy, i.e., minimizing provisioning cost and SLO violations, under various workload patterns.

Software cost estimation is a key process in project management. Estimations in the initial project phases are made with a lot of uncertainty that influences estimation accuracy which typically increases as the project progresses in time. Project data collected during the various project phases can be used in a progressive time-dependent fashion to train software cost estimation models. Our motivation is to reduce uncertainty and increase confidence based on the understanding of patterns of effort distributions in development phases of real-world projects. In this work, we study effort distributions and suggest a four-stage progressive software cost estimation model, adjusting the initial effort estimates during the development life-cycle based on newly available data. Initial estimates are reviewed on the basis of the experience gained as development progresses and as new information becomes available. The proposed model provides an early, a post-planning, a post-specifications, and a post-design estimate, while it uses industrial data from the ISBSG (R10) dataset. The results reveal emerging patterns of effort distributions and indicate that the model provides effective estimations and exhibits high explanatory value. Contributions in lessons learned and practical implications are also provided.

Purpose This paper aims to examine the connection between information system (IS) availability and operational risk losses and the capital requirements. As most businesses today become increasingly dependent on information technology (IT) services for continuous operations, IS availability is becoming more important for most industries. However, the banking sector has particular sector-specific concerns that go beyond the direct and indirect losses resulting from unavailability. According to the first pillar of the Basel II accord, IT outages in the banking sector lead to increased capital requirements and thus create an additional regulatory cost, over and above the direct and indirect costs of an outage. Design/methodology/approach A Bayesian belief network (BBN) with nodes representing causal factors has been used for identification of the factors with the greatest influence on IS availability, thus helping in investment decisions. Findings Using the BBN model for making IS availability-related decisions action (e.g. bringing a causal factor up to the best practice level), organization, according to the presented mapping table, would have less operational risk events related to IS availability. This would have direct impact by decreasing losses, related to those events, as well as to decrease the capital requirements, prescribed by the Basel II accord, for covering operational risk losses. Practical implications An institution using the proposed framework can use the mapping table to see which measures for improving IS availability will have a direct impact on operational risk events, thus improving operational risk management. Originality/value The authors mapped the factors causing unavailability of IS system to the rudimentary IT risk management framework implied by the Basel II regulations and, thus, established an otherwise absent link from the IT availability management to operational risk management according to the Basel II framework.

Appropriately defining and efficiently calculating similarities from large data sets are often essential in data mining, both for gaining understanding of data and generating processes and for building tractable representations. Given a set of objects and their correlations, we here rely on the premise that each object is characterized by its context, i.e., its correlations to the other objects. The similarity between two objects can then be expressed in terms of the similarity between their contexts. In this way, similarity pertains to the general notion that objects are similar if they are exchangeable in the data. We propose a scalable approach for calculating all relevant similarities among objects by relating them in a correlation graph that is transformed to a similarity graph. These graphs can express rich structural properties among objects. Specifically, we show that concepts—abstractions of objects—are constituted by groups of similar objects that can be discovered by clustering the objects in the similarity graph. These principles and methods are applicable in a wide range of fields and will be demonstrated here in three domains: computational linguistics, music, and molecular biology, where the numbers of objects and correlations range from small to very large.

Nowadays, many of the modern embedded applications such as vehicles and robots, interact with the environment and receive huge amount of data through various sensors such as cameras and radars. The challenge of processing large amount of data, within an acceptable performance, is solved by employing embedded systems that incorporate complementary attributes of CPUs and Graphics Processing Units (GPUs), i.e., sequential and parallel execution models. Component-based development (CBD) is a software engineering methodology that augments the applications development through reuse of software blocks known as components. In developing a CPU-GPU embedded application using CBD, allocation of components to different processing units of the platform is an important activity which can affect the overall performance of the system. In this context, there is also often the need to support and achieve run-time component allocation due to various factors and situations that can happen during system execution, such as switching off parts of the system for energy saving. In this paper, we provide a solution that dynamically allocates components using various system information such as the available resources (e.g., available GPU memory) and the software behavior (e.g., in terms of GPU memory usage). The novelty of our work is a formal allocation model that considers GPU system characteristics computed on-the-fly through software monitoring solutions. For the presentation and validation of our solution, we utilize an existing underwater robot demonstrator.