Service-Oriented Computing

Driven by the emerging business models (e.g., digital sales) and IT technologies (e.g., DevOps and Cloud computing), the architecture of software is shifting from monolithic to microservice rapidly. Benefit from microservice, software development, and delivery processes are accelerated significantly. However, along with many micro services running in the dynamic cloud environment with complex interactions, identifying and locating the abnormal services are extraordinarily difficult. This paper presents a novel system named “Microscope” to identify and locate the abnormal services with a ranked list of possible root causes in Micro-service environments. Without instrumenting the source code of micro services, Microscope can efficiently construct a service causal graph and infer the causes of performance problems in real time. Experimental evaluations in a micro-service benchmark environment show that Microscope achieves a good diagnosis result, i.e., 88% in precision and 80% in recall, which is higher than several state-of-the-art methods. Meanwhile, it has a good scalability to adapt to large-scale micro-service systems.

Microservice architectures are considered really promising to achieve devops in IT organizations, because they split applications into services that can be updated independently from each others. But to protect SLA (Service Level Agreement) properties when updating microservices, devops teams have to deal with complex and error-prone scripts of management operations. In this paper, we leverage an architecture-based approach to provide an easy and safe way to update microservices.

We present heuristics that help to identify suitable consumer-oriented parts of enterprise systems which could be re-engineered as microservices. Our approach assesses the key structural and behavioural properties common to both enterprise and microservice systems, as needed to guide a microservices discovery process and coherently assess restructuring recommendations. Building upon existing business object and system structural definitions, we present heuristics for two fundamental areas of microservice discovery, namely function splitting based on object subtypes (i.e., the lowest granularity of software based on structural properties) and functional splitting based on common execution fragments across software (i.e., the lowest granularity of software based on behavioural properties). A prototype analysis tool was developed based on the defined heuristics and experiments show that it can identify microservice designs which support multiple microservice characteristics, such as high cohesion, low coupling, high scalability, high availability, and processing efficiency while preserving coherent features of enterprise systems. In particular, we illustrate the usefulness of this new approach by conducting a case study based on customer management systems: SugarCRM and ChurchCRM.

Containerized microservices have become popular for building systems using simple, lightweight, loosely coupled services. Due to replacing the monolithic application with multiple microservices, inner function calls become inter-microservice communications, which increases the network pressure. However, the networking of containerized microservice built on the kernel that is inefficient. In this paper, we propose DockNet, a high-performance userspace networking solution for containerized microservices. We (1) leverage DPDK and customized LwIP as the high-performance data plane and TCP/IP stack, respectively. (2) introduce a master-slave threading model to decouple execution and management. (3) adopt namespace mechanism to control the access of microservices to data planes and employ timer-based rate limiters to achieve performance isolation. (4) construct fast channels between partner microservices to improve network performance further. In our various experiments, DockNet shows over $$4.2\times , 4.3\times , 5.5\times $$ of higher performance compared with existing networking solutions - kernel bridge, Open vSwitch and SR-IOV, respectively.

Microservice APIs represent the client perspective on microservice-based software architecture design and related practices. Major issues in API design concern the quality aspects of the API. However, it is not well understood today what the established practices related to those quality aspects are, how these practices are related, and what the major decision drivers are. This leads to great uncertainty in the design process. In this paper, we report on a qualitative, in-depth study of 31 widely used APIs plus 24 API specifications, standards, and technologies. In our study we identified six recurring architectural design decisions in two API design contexts with a total of 40 decision options and a total of 47 decision drivers. We modelled our findings in a formal, reusable architectural decision model. We measured the uncertainty in the resulting design space with and without use of our model, and found that a substantial uncertainty reduction can be potentially achieved by applying our model.

To achieve on-time completion of time-constrained business cloud workflows, a large number of parallel cloud workflow instances need to be constantly monitored so that temporal violations (namely intermediate runtime delays) can be detected and handled timely. Over the last few years, many strategies have been proposed but they are not adaptive enough to capture the dynamic behaviors of business cloud workflows. In this paper, we introduce the idea of “adaptiveness” into our strategy design. Specifically, we first present an adaptive temporal checkpoint selection strategy where the time intervals between checkpoints are adaptively determined at runtime, and then propose a matching temporal violation handling strategy which can determine the required lifecycle of cloud services. The evaluation results demonstrate that our adaptive strategy can achieve both higher efficiency and better cost effectiveness compared with conventional strategies.

Business process modeling is an integral task needed for efficient running of business operations. Often process models remain buried in unstructured documents as images or screenshots. Such embedded process model images may become quickly obsolete as the underlying business process evolves. Thus, there is value in digitizing the unstructured images. We propose a novel automated solution to transform a process model image into the standard Business Process Model and Notation (BPMN) format. Our deep-learning based approach performs well in practice achieving good precision and recall.

The importance and adoption of Blockchain to support secure and trusted collaborations between businesses continues to grow. In today’s practice, most Blockchain smart contracts (which capture the business processing logic) are written primarily by software developers. To enable widespread adoption of Blockchain, business analysts and subject matter experts will need to have direct access to the smart contract logic, including the abilities to understand, modify, and create substantial portions of that logic. This paper describes a fully functioning framework and system for specifying and executing smart contracts in which the core logic is specified by a controlled English, business-level rules language. The framework includes a browser-based smart editor for rules; a parser generator that enables substantial variation in the rules syntax; code generation that maps to a RETE based rules engine; and execution of the rules in either on-chain (using Hyperledger Fabric) or off-chain modes. The paper describes the rules framework and possible extensions, and identifies key aspects of Blockchain that impact the implementation.

In Social Internet of Things (SIoT) environments, a large number of users and Internet of Things (IoT) based devices are connected to each other, so that they can share SIoT-based services. IoT-based devices establish social relations with each other according to the social relations of their owners in Online Social Networks (OSNs). In such an environment, a big challenge is how to provide trustworthy service evaluation. Currently, the prevalent trust management mechanisms consider QoS-based trust and social-relation based trust mechanisms in evaluating the trustworthiness of service providers. However, the existing trust management mechanisms in SIoT environments do not consider the different contexts of trust. Therefore, dishonest SIoT devices, based on their owners’ social relations, can succeed in advertising low-quality services or exploiting maliciously provided services. In this paper, we first propose three contexts of trust in SIoT environments including the status and environment (time and location) of devices, and the types of tasks. Then, we propose a novel Mutual Context-aware Trustworthy Service Evaluation (MCTSE) model. The experiments demonstrate that our proposed contextual trust evaluation model can effectively differentiate honest and dishonest devices and provide a high success rate in selecting the most trustworthy services and providing high resilience against different attacks from dishonest devices.

In this paper, we introduce, design and develop Cloudchain, a blockchain-based cloud federation, to enable cloud service providers to trade their computing resources through smart contracts. Traditional cloud federations have strict challenges that might hinder the members’ motivation to participate in, such as forming stable coalitions with long-term commitments, participants’ trustworthiness, shared revenue, and security of the managed data and services. Cloudchain provides a fully distributed structure over the public Ethereum network to overcome these issues. Three types of contracts are defined where cloud providers can register themselves, create a profile and list of their transactions, and initiate a request for a service. We further design a dynamic differential game among the Cloudchain members, with roles of cloud service requesters and suppliers, to maximize their profit. Within this paradigm, providers engage in coopetitions (i.e., cooperative competitions) with each other while their service demand is dynamically changing based on two variables of gas price and reputation value. We implemented Cloudchain and simulated the differential game using Solidity and Web3.js for five cloud providers during 100 days. The results showed that cloud providers who request services achieve higher profitability through Cloudchain compared to those providers that supply these requests. Meanwhile, spending high gas price is not economically appealing for cloud requesters with a high number of requests, and fairly cheaper prices might cause some delays in their transactions during the network peak times. The best strategy for cloud suppliers was found to be gradually increasing their reputation, especially when the requesters’ demand is not significantly impacted by the reputation value.

Account payables are amount owed to vendors for goods and services delivered to a company. Vendors raise invoices which go through several processing steps before they are paid by a company. Companies have contractual obligations with vendors for paying the invoices within a stipulated time. Invoices that exceed this time attract penalty and affect vendor satisfaction to work with the company. It is very critical for large firms dealing with thousands of vendors for their day to day operations to meet the service level agreements with vendors to avoid penalties. Any assistance for practitioners, warning them of potential invoices that can breach the service level agreements, can help them in minimizing the penalties. In this research, we model the problem of identifying delayed invoices as a supervised classification task. There are three characteristics of this problem which are challenging from a classification perspective: (i) the status of an invoice is affected by other invoices that are simultaneously being processed, as there are limited resources to process the huge volume of invoices, (ii) feature engineering to capture the temporal aspect of the invoice and having the optimal representation of the multiple data entries created per invoice, and (iii) the number of paid late invoices are much smaller in percentage compared to paid on time invoices in the training data set, hence the classes are imbalanced. The results obtained by training an ensemble of classifiers show that penalties can be avoided on more than 82% of the invoices which are currently being penalized.

Modern companies execute business processes to deliver products and services, whose enactment requires to adhere to laws and regulations. Compliance checking is the task of identifying potential violations of such requirements prior to process execution. Traditional approaches to compliance checking employ formal verification techniques (e.g., model checking) to identify which process paths in a process model may lead to violations. However, this diagnostics is, in most of the cases, not rich enough for the user to understand how the process model should be changed to solve the violations. In this paper, we present an approach based on finite-state automata manipulation to identify the specific process activities that are responsible to cause violations and, in some cases, suggest reparative actions to be applied to the process model to solve the violations. We show that our approach can be expressed as a planning problem in Artificial Intelligence, which can be efficiently solved by state-of-the-art planners. We report experimental results using synthetic case studies of increasing complexity to show the scalability of our approach.

With the generalization of the Cloud, software providers can distribute their software as a service without investing in large infrastructure. However, without an effective resource allocation method, their operation cost can grow quickly, hindering the profitability of the service. This is the case for BPM as a Service providers that want to handle hundreds of customers with a given quality of service. Since there are variations in the needed load and in the number of users of the service, the allocation and scheduling methods must be able to adjust the cloud resource quantity and size, and the distribution of customers on these resources. In this paper, we present a cost optimization model and an heuristic based on genetic algorithms to adjust resource allocation to the needs of a set of customers with varying BPM task throughput. Experimentations using realistic customer loads and cloud resources capacities show the gain of these methods compared to previous approaches. Results show that using our algorithm on split groups of customers can provide even better results.

The interest in processing data events under stringent time constraints as they arrive has led to the emergence of architecture and engines for data stream processing. Edge computing, initially designed to minimize the latency of content delivered to mobile devices, can be used for executing certain stream processing operations. Moving operators from cloud to edge, however, is challenging as operator-placement decisions must consider the application requirements and the network capabilities. In this work, we introduce strategies to create placement configurations for data stream processing applications whose operator topologies follow series parallel graphs. We consider the operator characteristics and requirements to improve the response time of such applications. Results show that our strategies can improve the response time in up to 50% for application graphs comprising multiple forks and joins while transferring less data and better using the resources.

In mobile edge computing, edge servers are geographically distributed around base stations placed near end-users to provide highly accessible and efficient computing capacities and services. In the mobile edge computing environment, a service provider can deploy its service on hired edge servers to reduce end-to-end service delays experienced by its end-users allocated to those edge servers. An optimal deployment must maximize the number of allocated end-users and minimize the number of hired edge servers while ensuring the required quality of service for end-users. In this paper, we model the edge user allocation (EUA) problem as a bin packing problem, and introduce a novel, optimal approach to solving the EUA problem based on the Lexicographic Goal Programming technique. We have conducted three series of experiments to evaluate the proposed approach against two representative baseline approaches. Experimental results show that our approach significantly outperforms the other two approaches.

As the Internet of Things (IoT) technologies permeate people’s daily lives, the sheer number of IoT applications has been developed to provide a wide range of services. Among all, real-time IoT services start to draw increasing attentions. Conventionally, cloud plays the role as the service provider in IoT but is no longer considered as the rational option for the real-time services due to service transmission latency and communication overhead. Therefore, we propose a novel rate-adaptive fog service delivery platform, namely RA-FSD, aiming at real-time service provisioning and network utility maximization (NUM) of the underlying IoT resources based on the newly emerged fog computing paradigm. The platform leverages fog nodes as either fog service provider to offer timely services for end users, or service intermediaries to help track network conditions and mitigate communication overhead. By doing so, service consumers would always benefit from the fact that services produced by IoT applications are in their proximity and thus delivered to destination in a prompt manner. A service rate-adaptive algorithm is also developed as the key component of the RA-FSD platform to handle the abrupt changes happened in IoT network, dynamically adjust service delivery rate based on the network condition while retaining satisfactory Quality of Service (QoS) to each service consumer, and support both elastic and inelastic network services from heterogeneous IoT applications.

Existing event processing models require defining events in details beforehand. It is thus challenging to handle uncertainty associated with various sensor streams having dynamic interventions and correlations. In this paper, we improve our previous service abstraction which can increase the value density of primitive sensor streams in two aspects. To deal with the uncertainty, we add declarative rules in our service abstraction for adaptively generating events from different sensor streams that reflect various external stimuli. For extracting events dynamically, we utilize the correlation analysis method to treat events as variations of correlations. This paper reports the tryout use of our approach in Chinese power grid for detecting abnormal situations of power quality.

A typical complex event processing (CEP) service is composed by a set of operators organized as a directed acyclic graph. This kind of service is usually used to handle large amounts of real-time data. Meanwhile, edge computing has been widely accepted as a new paradigm to improve the QoS of deployed services by making the services closer to the data. Thus, the response time, which is a crucial QoS metric for CEP services, can be significantly reduced by deploying CEP services on the edge network. However, it is often unlikely for a single node of the edge network to host all operators of a CEP service due to the limited computing resources. Therefore, it is desirable for a CEP service to place its operators on different nodes of the edge network to keep the response time low, especially when the input rate of the CEP service significantly increases. In this paper, we reduce the average response time of CEP services by deploying the operators on the edge nodes dynamically according to the predicted response time of CEP services. Specifically, we first propose a system model to capture the response time of the CEP services, based on which we formulate the problem of the optimal placement of CEP operators in the edge network. We then propose an algorithm that predicts the response time of CEP services and deploys the operators on the edge nodes with the minimum predicted delay. A simulation-based evaluation demonstrates that, compared with two state-of-the art algorithms, our algorithm can reduce the total response time by 33 $$\%$$ and 45 $$\%$$ on average, respectively.

The Internet of Things (IoT) pervades more and more aspects of our lives and often involves many types of smart connected objects and devices. User’s IoT environment changes dynamically, e.g., due to the mobility of the user and devices. Users can fully benefit from the IoT only when they can effortlessly interact with it. To accomplish this in a dynamic and heterogenous environment, we make use of Emergent Configurations (ECs), which consist of a set of things that connect and cooperate temporarily through their functionalities, applications, and services, to achieve a user goal. In this paper, we: (i) present the IoT-FED architectural approach to enable the automated formation and enactment of ECs. IoT-FED exploits heterogeneous and independently developed things, IoT services, and applications which are modeled as Domain Objects (DOs), a service-based formalism. Additionally, we (ii) discuss the prototype we developed and the experiments run in our IoT lab, for validation purposes.

Cloud based robotic services can be adopted for emergency management in smart factory. When multiple robots work collaboratively in such system, optimal resource allocation for executing the tasks of robotic services becomes a challenging problem due to the heterogeneity and energy consumption of resources. Since the tasks of multi-robotic services are inter-dependent, the inconvenience of data exchange between local robots and distant Cloud can significantly degrade the quality of service. Therefore, in this paper, we jointly address the energy consumption and service delay minimization problem while allocating resources in proximate Cloud (Cloudlet) based multi-robot systems for emergency management service in smart factory. A multi-objective evolutionary approach, NSGA-II algorithm is applied to solve this constrained multi-objective optimization problem. We augment the NSGA-II algorithm by defining a new chromosome structure, presorted initial population, mutation operator and selection of minimum distant solution from the non-dominated front to the origin while crossing over the chromosomes. The experimental results on the basis of synthetic data demonstrate that our approach performs significantly well compared to benchmark NSGA-II.

IoT offers a large number of services from different providers. These services frequently need to be composed to provide novel applications. Current work in IoT service composition can be classified as conversation-based or interface-based. Conversation-based approaches need the manual definition of service plans, which is not feasible in IoT because of the large scale. Interface-based approaches use planning to automate the composition process. Such automation avoids human intervention, but some incorrect services can appear in the discovered plans. The efficiency of these approaches is poor because they perform intensive search in large spaces. This paper proposes a model for service composition with minimal human intervention using consumers’ feedback. Results show that our model outperforms its competitors.

With the growing popularity of mobile devices, a new paradigm called mobile crowdsourcing emerged in the recent years. Mobile users with restricted computational capability and sensing ability are now able to conduct complex tasks with the help of other users in the same mobile crowdsourcing system. In this paper, we consider the mobile crowdsourcing system model based on the spontaneously-formed mobile social networks (MSNs). We introduce two crowdsourcing task scheduling problems under this system model, with one problem aiming to minimize the operating cost of some crowdsourcing tasks and the other focusing on minimizing the overall completion time of tasks belonging to the same project. Correspondingly, under offline settings, we propose an optimal algorithm and an approximation algorithm for these two problems respectively. The optimality and the approximation ratio are analyzed accordingly. Based on these two algorithms, we further design two online algorithms to deal with the problems under online settings and their competitive ratios are computed. Finally, we verify the effectiveness and efficiency of the proposed methods through extensive numerical experiments on synthetic datasets.

Ticket assignment/dispatch is a crucial part of service delivery business with lot of scope for automation and optimization. In this paper, we present an end-to-end automated helpdesk email ticket assignment system, which is also offered as a service. The objective of the system is to determine the nature of the problem mentioned in an incoming email ticket and then automatically dispatch it to an appropriate resolver group (or team) for resolution.The proposed system uses an ensemble classifier augmented with a configurable rule engine. While design of classifier that is accurate is one of the main challenges, we also need to address the need of designing a system that is robust and adaptive to changing business needs. We discuss some of the main design challenges associated with email ticket assignment automation and how we solve them. The design decisions for our system are driven by high accuracy, coverage, business continuity, scalability and optimal usage of computational resources.Our system has been deployed in production of three major service providers and currently assigning over 40,000 emails per month, on an average, with an accuracy close to 90% and covering at least 90% of email tickets. This translates to achieving human-level accuracy and results in a net saving of about 23000 man-hours of effort per annum.

We propose a new framework for crowdsourcing energy services from Internet-of-Things devices. We introduce a new crowdsourced energy as a service and energy-related quality model considering spatiotemporal aspects. We describe a new temporal composition algorithm to compose energy services to satisfy a user’s energy requirement. The temporal composition algorithm is a variation of fractional knapsack algorithm. We conduct preliminary experiments to demonstrate the performance and effectiveness of our approach.

We consider the scene analysis as a service composition problem. A social-sensor cloud services composition model is proposed for the scene analysis. Our proposed model selects and composes social-sensor cloud services based on the user queries. Textual features of the social-sensor cloud services, i.e., description, comments, and meta-data of the social media images are used to reconstruct a scene. Our key contribution is an efficient and real-time composition of related images for scene analysis relying on meta-data and related posted information. Analytical results demonstrate the performance of the proposed model.

With the rapid proliferation of mobile devices, Mobile Crowdsensing (MCS) has become an efficient way to ubiquitously sense and share environment data. Due to the openness of MCS, sensors and workers are of different qualities. Low quality sensors and workers may yield low sensing quality. Thus it is important to infer workers’ qualities and seek a valid task assignment with enough total qualities for MCS. To solve the quality inference problem, we adopt truth inference methods to iteratively infer workers’ qualities. This paper also proposes an task assignment problem called quality-bounded task assignment with redundancy constraint (QTAR) based on truth inference. We prove that QTAR is NP-complete and propose a $$ (2+\epsilon ) $$ - approximation algorithm QTA for QTAR. Finally, experiments conducted on real dataset prove the efficiency and effectiveness of the algorithms.

Knowledge acquisition and exchange are generally crucial yet costly for both businesses and individuals, especially when the knowledge concerns various areas. Question Answering Communities offer an opportunity for sharing knowledge at a low cost, where communities users, many of whom are domain experts, can potentially provide high-quality solutions to a given problem. In this paper, we propose a framework for finding experts across multiple collaborative networks. We employ the recent techniques of tree-guided learning (via tensor decomposition), and matrix factorization to explore user expertise from past voted posts. Tensor decomposition enables to leverage the latent expertise of users, and the posts and related tags help identify the related areas. The final result is an expertise score for every user on every knowledge area. We experiment on Stack Exchange Networks, a set of question answering websites on different topics with a huge group of users and posts. Experiments show our proposed approach produces steady and premium outputs.

With broad application of Web service technology, many users look for applicable Web services to construct their target application quickly or do some further research. Github, as a treasury including a variety of software programs, provides functional code modules for those in need, which has become their characteristic service. However, tremendous Web services have been developed all the time which increase the difficulty to find the target or interested services for users. Service recommendation has become of practical importance. There is few studies in the personalized repository recommendation of Github. In this paper, we present a general framework of PNCF, a preference-based neural collaborative filtering recommender model, and develop the instantiation of PNCF framework in Github repository recommendation with language preference called LR-PNCF. We use a neural network to capture the non-linear user-repository relationships and obtain abstract data representation from sparse vectors. Comprehensive experiments conducted on a real world dataset demonstrate the effectiveness of the proposed approach.

Explosive growth in the number of mobile applications (apps) makes it difficult for users to find relevant apps. Therefore, it is an urgent task to recommend desired apps for users. Traditional approaches focus on exploiting the context information, user’s interest, privacy, security and other features for app recommendation. Most of them do not consider heterogeneous information network (HIN) in the scenario of mobile app recommendation. HIN contains rich structure and semantic information, and it can satisfy various requirements of users and generate better recommendation results. In this paper, we propose a Weighted Meta-Graph based approach for app Recommendation, called WMGRec, on HIN. Specifically, we firstly introduce the concept of weighted meta-graph, which not only distinguishes different rating scores to depict the subtle semantics but also utilizes meta-graph to capture complex semantics. And then, we apply weighted meta-graph to measure the semantic similarity between users and apps. Furthermore, we leverage non-negative matrix factorization on user-app similarity matrix to obtain user latent features and app latent features. Finally, the concatenated user and app latent features are fed into the factorization machine & deep neural network model to learn the higher-order interactions and get the final prediction score. Extensive experiments conducted on two real-world datasets validate the effectiveness of the proposed approach compared to state-of-the-art recommendation algorithms.

Temporal information has been proved to be an important factor to recommender systems. Both of user behaviors and QoS performance of services are time-sensitive, especially in dynamic cloud environment. Furthermore, due to the data sparsity problem, it is still difficult for existing recommendation methods to get the similarity relationships between services or users well. In view of these challenges, in this paper, we propose a temporal-sparsity aware service recommendation method based on hybrid collaborative filtering (CF) techniques. Specifically, temporal influence is considered into classical neighborhood-based CF model by distinguishing temporal QoS metrics from stable QoS metrics. To deal with the sparsity problem, a time-aware latent factor model based on a tensor decomposition model is applied to mine the temporal similarity between services. Finally, experiments are designed and conducted to validate the effectiveness of our proposal.

With the overwhelming increase of web services on the Internet, how to accurately perform QoS prediction has played a key role in service recommendation. Recently, three kinds of approaches have been presented on service QoS prediction based on collaborative filtering (CF), including user-intensive, service-intensive and their combination. However, the deficiency of current approaches is that all of the services invoked by target user (or all of the users who invoked target service) are applied to calculate average QoS, without the reduction to those dissimilar with target service (or target user). In this paper, we propose a reinforced collaborative filtering approach, where both similar users and services are integrally considered into a singleton CF. The experiments are conducted on a large-scale dataset called WS-DREAM, involving 5,825 real-world Web services in 73 countries and 339 service users in 30 countries. The experimental results demonstrate that our approach for QoS prediction outperforms the competing approaches.

Service discovery and recommendation is playing increasingly important role, as more and more reusable web services are published onto the Internet. Existing methods typically recommend either individual services, or multiple services without their interconnections. In contrast, this research aims to mine service usage history and extract units of work (UoWs) comprising a collection of services chained together through intermediate components. A novel technique is proposed in this paper to study how services collaborated, or could collaborate, in the form of reusable UoWs to serve various workflows (i.e., mashups), based on an evolving service social network. Upon receiving a scientific workflow request, a recommend-as-you-go algorithm simulates how human minds work and relies on a sliding aggressiveness gauge to incrementally recommend context-aware UoWs. In this way, we hope to move one step further toward automatic service composition. Extensive experiments on the real-world datasets demonstrate the effectiveness and efficiency of the UoW-oriented service recommendation approach.

We propose a mobile crowdsourced sensors selection approach to improve the journey planning service especially in areas where no wireless or vehicular sensors are available. We develop a location estimation model of journey services based on an unsupervised learning model to select and cluster the right mobile crowdsourced sensors that are accurately mapped to the right journey service. In our model, the mobile crowdsourced sensors trajectories are clustered based on common features such as speed and direction. Experimental results demonstrate that the proposed framework is efficient in selecting the right crowdsourced sensors.

Most traditional business process recommendation methods cannot deal with complex structures such as interacting loops, and they cannot handle large complex datasets with a great quantity of processes and activities. To address these issues, RLRecommender, a method based on representation learning, is proposed. RLRecommender extracts three kinds of relation sets from the models, both activities and relations between them are projected into a continuous low-dimensional space, and proper activity nodes are recommended by comparing the distances in the space. The experimental results show that our method not only outperforms other baselines on small dataset, but also performs effectively on large dataset.

IT service support agents are trained on knowledge sources with large volumes of domain-specific documents, including product manuals and troubleshooting contents. Self-assist applications, such as search and support chat-bots must integrate such knowledge in order to conduct effective user interactions. In particular, the very large volume of domain-specific terms referenced in training documents must be accurately identified and qualified for relevance to specific context of support actions. We propose a weakly-supervised approach for extraction of key terms from IT support documents. The approach integrates domain knowledge to refine the extraction results. Our approach obviates the need for extensive expert work creating manual annotation and dictionary collection, as typically required in traditional supervised solutions, as well as the limited accuracy obtained in unsupervised methods. Results show that domain knowledge based refinement helps improve the overall accuracy of mined key terms by 25–30%.

The issue of node ranking in complex networks is a classical problem that has obtained much attention over the past few decades, and a great variety of methods have consequently been developed. These proposed methods can be roughly categorized into the global and local methods. The global methods are usually time-consuming and the local methods may be inaccurate. In this paper, we propose a novel semi-local algorithm ASLA (Adaptive Semi-Local Algorithm) that seeks a tradeoff between the time efficiency and the ranking accuracy to overcome the limitations of the global and local methods. ASLA is able to adaptively determine the potential influence scope for each node. Then, the influence value of each node is calculated based on such a personalized influence scope. Finally, all the nodes are ranked according to their influence values. To evaluate the performance of ASLA, we have conducted extensive experiments on both synthetic networks and real-world networks, with the results demonstrating that ASLA is not only more efficient than the global methods but also more accurate than the local methods.

In recent years, mobile crowdsourcing has been integrated into people’s lives. A variety of mobile crowdsourcing services have emerged and been widely used, such as Gigwalk, Foursquare, and Uber. Due to the uncertainty of task distribution and workers’ trajectory, as well as diverse worker interests and capabilities, it is crucial to effectively predict the mobile workers’ trajectories such that they are willing to get to the location and perform their tasks with as little travel and time cost as possible. In this paper, we propose a context-sensitive prediction approach for workers’ moving path in mobile crowdsourcing services. We predict the upcoming location of workers through movement rules, real-time perception of workers’ moving path and contexts when assigning spatial tasks on a crowdsourcing platform, thereby pushing a task to the workers who will enter the region within the deadline of the task. Our location prediction method can avoid workers’ extra cost such as time and charges in performing tasks. The analysis and simulation experiments based on real data sets show that this method can effectively predict the location of a worker and achieve better results in task assignment and completion.

Industry-scale context-aware processes typically manifest a large number of variants during their execution. Being able to predict the performance of a partially executed process instance (in terms of cost, time or customer satisfaction) can be particularly useful. Such predictions can help in permitting interventions to improve matters for instances that appear likely to perform poorly. This paper proposes an approach for leveraging the process context, process state, and process goals to obtain such predictions.

At present, most publish/subscribe systems assume that all participants have the same Quality of Service (QoS) requirements. However, in many real-world IoT service scenarios, different users may have different delay requirements. How to provide differentiated services has become an urgent problem. The rise of Software Defined Networking (SDN) provides endless possibilities for meeting customized services due to greater programmability. In this paper, we first propose two new methods to predict the queuing delay of switches. One is an improvement of the traditional Random Early Detection (RED) algorithm; the other is a machine learning method using the eXtreme Gradient Boosting (XGBoost) model. Then we describe an SDN-like publish/subscribe system architecture and priority queues supported by OpenFlow switches to realize differentiated services. In order to guarantee QoS, we present a two-layer queue management mechanism based on user requirements. In the end, we compare our delay prediction methods with the RED method and verify the effectiveness of the two-layer queue management mechanism. Experimental results show that our solution is effective.

In the current Web Service Description Language (WSDL), only the interface information of a web service is provided without any indication on its behavior logic. Naturally, it is difficult for the service user and developer to achieve a shared understanding of the service behavior through such a description. A particular challenge is how to make explicit the various behavior assumptions and restrictions of a service (for the user), and make sure that the service implementation conforms to them (for the developer). In order to improve the behavior conformance of services, in this paper we propose a constraint-based model-driven testing approach for web services. In our approach, constraints are introduced in an extended WSDL, called CxWSDL, to formally and explicitly express the implicit restrictions and assumptions on the behavior of web services, and then the predefined constraints are used to derive test cases in a model-driven manner to test the service implementation’s conformance to these behavior constraints from the user’s perspective. We have conducted an empirical study with three real-life web services as subject programs, and the experimental results have shown that our approach can effectively validate the service’s conformance to the behavior constraints.

A service cloud could improve its QoS (Quality of Service) by partitioning jobs into multiple tasks and processing those tasks in parallel. In contrast to processing all jobs with the same degree of parallelism (DOP), dividing jobs into different groups and processing them with varying DOPs may achieve better performance results, especially focusing on those jobs which have a greater impact on performance of service clouds. In this paper, we describe a novel differentiated DOP policy, which divides jobs into several groups identified by jobs’ service time and sets proper DOPs for different groups of jobs. Then, we propose a parallel multi-queue and multi-station analytical model for service clouds with our differentiated DOP policy, to predict important performance metrics. Thus this model can guide cloud providers to determine optimal DOPs and resource allocation schemes for different groups to improve the total QoS of a service cloud. We also present a new metric, called Optimized Performance of Groups (OPG), to quantify the level of performance optimization of every group. The objective is to maximize the minimum OPG to ensure OPG within a certain range, thereby enforcing a fair trade-off between all groups. Through extensive experiments, we validate the effectiveness of the proposed differentiated DOP policy and analytical model.

Estimating future runtime performance and cost is an essential task for Chief Information Officers in deciding whether to adopt a Cloud-based system. Benchmarking and simulation are two techniques that have long been practiced towards reliable estimation. Benchmarking involves (potentially) high cost and time consumption, but oftentimes yields more reliable estimates than simulation, while the simulation is much cheaper and faster than benchmarking, but less reliable. In order to deal with this dichotomy, we propose a complementary approach to estimating the performance of Cloud-based systems, whereby performance estimates can be obtained in a fast, inexpensive, and also reliable way. In this approach, the ontological concepts of a benchmark model, whose benchmark results have already been obtained, are mapped into those of a simulation model, while the mismatches and similarities between the two models are taken care of, through measures of similarity between the two. This ontology-driven construction of simulation models is intended not only to yield more reliable simulation results but also to help better explain why the simulation results may, or may not, be reliable. To validate our complementary approach, simulation models are constructed using CloudSim, and the simulation results are compared against the corresponding benchmark results, by using our prototype tool, collected from Amazon Web Service (AWS) and Google Compute Engine (GCE) by using the Yahoo! Cloud Serving Benchmark (YCSB) tool. These experiments show that the simulation results show about 90% accuracy with respect to the benchmark results, and additionally we feel we could better explain why this happens.

Collaborative filtering (CF) has been widely used in quality of service (QoS) prediction. However, most of traditional CF-based methods always suffer from overestimation of similarity computation and invalid neighbors. To address these problems, we propose a two-phase QoS prediction approach based on restricted Boltzmann machine (RBM). In the first phase, we propose an RBM-based approach to predict missing QoS values for invalid neighbors, which can identify similar neighbors with high accuracy. In the second phase, we propose a user-based CF method to predict, which utilizes user similar neighbors. Experimental results conducted in a real-world dataset show that our approaches can produce superior prediction accuracy and are not sensitive to parameter settings.

Web-APIs enable cross-organizational functionality integration over the Web and thus are the foundation of modern distributed service-based systems. However, despite the rapid increase in the number of Web-APIs available on the Internet, the discovery and uptake of appropriate Web-APIs by businesses on a Web scale is still a great challenge. One of the main reasons is that Web-APIs registered on directories such as ProgrammableWeb.com are in general isolated, as they are registered by diverse providers independently and progressively. In this paper, we present a method for analyzing the Web-API ecosystem and propose a complex-network-based approach for building an evolving social network for Web APIs. We conduct our analysis in two phases: First, from the complex network perspective, we investigate mashups and Web-APIs interactions and analyze the Web-API popularity distribution using the popular ProgrammbleWeb dataset. Second, we quantitatively measure the Preferential Attachment mechanism which is a key driver of an evolving network. Based on our analysis, we propose an approach to construct an evolving Web-API social network based on the theoretical procedure of the Barabási-Albert complex network model. Results presented in this work will not only provide insight into the topology of the Web-API ecosystems but also serve as a practical guide for designing an evolving-network-based solution for service discovery.

Users within a limited geographic area can form service-sharing communities using the services deployed on their mobile devices. Creating Quality of Service (QoS) optimal service compositions in such decentralised and dynamic environments is challenging because of the service providers’ mobility and the inherent dynamism in the available services. Existing proposals for mobile environments either use template-matching composition or require a-priori knowledge about the QoS objectives’ weights, which limits the composition’s flexibility in such environments. This paper presents a stigmergic-based approach to model the decentralised, flexible and dynamic service interactions of providers in a mobile environment. A nature-inspired optimisation mechanism is used to approximate the set of QoS optimal compositions that result from these interactions. To facilitate adaptation of the composite during execution, we introduce a procedure that encourages the exploration of service composition configurations that emerge as a result of providers’ mobility. We evaluate the performance of the proposed approach with a no-adaptation variant, a Dijkstra-based, a Greedy and a Random approach. The results show that the proposed approach can obtain superior solutions compared with current optimisation methods for flexible service composition in mobile environments at the cost of increased overhead.

The migration of legacy software systems to Service Oriented Architectures (SOA) has become a mainstream trend for modernizing enterprise software systems. A key step in SOA migration is the identification of services in the target application, but it is a challenging one to the extent that the potential services (1) embody reusable functionalities, (2) can be developed in a cost-effective manner, and (3) should be easy to maintain. In this paper, we report on state of the practice of SOA migration in industry. We surveyed 45 practitioners of legacy-to-SOA migration to understand how migration, in general, and service identification (SI), in particular are done. Key findings include: (1) reducing maintenance costs is a key driver in SOA migration, (2) domain knowledge and source code of legacy applications are most often used respectively in a hybrid top-down and bottom-up approach for SI, (3) industrial SI methods focus on domain services–as opposed to technical services, (4) there is very little automation of SI in industry, and (5) RESTful services and microservices are the most frequent target architectures. We conclude with a set of recommendations and best practices.

Recently, more and more users outsource their job executions to service clouds. To reduce the costs and risks, many service providers purchase on-demand instances from IaaS clouds to provide services elastically. For maximizing social welfare, service providers need effective approaches to optimally purchase IaaS instances and schedule parallel jobs which have soft deadline, according to the valuations reported by users. In order to address the challenges such as NP-hardness and possible misreports of users, we design an auction-style randomized mechanism for the instance purchasing as well as job scheduling and pricing problem in service clouds. This mechanism can achieve an approximately optimal social welfare while scheduling jobs in a way without preemption. Many critical properties can be guaranteed simultaneously by our mechanism, including truthfulness in expectation, computational efficiency and individual rationality. Both the theoretical analysis and the extensive simulations based on synthetic data and real-world job traces validate the effectiveness of our mechanism on social welfare maximization.

We propose a novel service mining framework to personalize services in an IoT based smart home. We describe a new technique based on the concept of convenience to discover periodic composite IoT services to suit the smart home occupant’s convenience needs. The key features of convenience is the ability to model the spatio-temporal aspects as occupants move in time and space within the smart home. We propose a novel framework for the transient composition of spatio-temporal IoT service. We design two strategies to prune non-promising compositions. Specifically, a significance model is proposed to prune insignificant composite IoT services. We describe a spatio-temporal proximity technique to prune loosely correlated composite IoT services. A periodic composite IoT service model is proposed to model the regularity of composite IoT services occurring at a certain location in a given time interval. The experimental results on real datasets show the efficiency and effectiveness of our proposed approach.

Mashups involve the collaboration of multiple developers to build Web applications out of pre-existing APIs. A large body of research focused on recommending APIs for mashups. However, very few contributions looked at recommending developers. In this paper, we propose CrowdMashup, a crowdsourcing approach for mashup teams recommendation. We analyze online developer communities and API directories to infer developers’ interests in APIs through natural language processing. We predict missing interest values using the alternating least square method for collaborative filtering. We also model interactions (comments and replies) among developers as a weighted undirected graph and introduce a sociometric to identify socially related developers. We propose an algorithm, based on the concept of cliques in graph theory, that combines developers’ skills and sociometric to recommend efficient and balanced teams. We describe a prototype implementation and conduct extensive experiments on real-world data and APIs to evaluate our approach.

Contemporary approaches for automated web service composition mostly deal with static web services. The underlying assumption here is that the web services participating in resolving a query are static and thereby, their functional and non-functional parameters change very infrequently or do not change at all. However, in reality, this assumption does not hold. New services are added to the repository, existing unpopular services are removed from the repository, service interfaces change due to changes in the specification. Classical service composition approaches therefore fall short to handle the dynamic behavior of a web service during composition. In this paper, we present a stochastic model of the web service composition problem to capture the dynamic behavior of web services from the functional perspective. We present experimental results on the ICEBE-2005 benchmarks to show the effectiveness of our proposed methods.

Cloud computing projects require the design of a so-called Cloud Solution, which is an architectural blueprint for a particular cloud environment. A cloud solution defines the hosting infrastructure (servers, VMs, etc.), software stack, and services such as network, backup, disaster recovery, management, etc. The design of a cloud solution needs to consider existing client environments and future environment’s requirements, and at the same time comply with the cloud provider’s portfolio and limitations. As such, the design of enterprise cloud solutions is a very complex and challenging problem. In this paper, we present a novel framework for provider-side cloud solution design based on model-driven and formal methods that facilitates the job of automated solution generation, starting from client requirements and resulting in a complete and correct cloud solution. We present a set of novel methods and a tool, called COOL, which implements the method and is used in production in a large Cloud service provider.

Key challenges in managing healthcare applications lie in the area of compliance of the deployment environments and the usage of hybrid clouds. Our approach, as reported in this paper, utilizes two innovative concepts: compliance conformance validation and environment reconstruction supported by a Platform as a Service (PaaS) environment performing healthcare application automated migrations in hybrid clouds. We show how the migration process is conducted with dynamic reconstruction of the application dependencies on the PaaS services. For system administrators, this approach can lead to significant time savings on migrations to compliant environments. Implementation details and experimental results are presented to validate our methodology.

E-commerce businesses compete in the market by conducting marketing strategies consisting of four aspects: customers, products, marketplaces and intermediaries. One of the widely-used marketing strategies, called Lucky Money, is capable of encouraging customers to buy products from marketplaces. However, the amount of luck money for each customer is usually randomly determined or even manually determined and cannot fully achieve the business objectives. This paper proposes a machine-learning based lucky money determination approach, called DAliM, for e-commerce businesses to achieve their desired goals. We implement DAliM for the “Double 11 Global Shopping Festival 2017” initiated by Alibaba Group and evaluate it using a few hundred million real customers from all over the world. The experimental results demonstrate that our method manages to decrease the lucky money spent by 41.71% and increase the final purchase rate by 24.94% compared to the state-of-the-art baseline.

Pervasive computing promotes the integration of smart electronic devices in our living and working spaces in order to provide new, advanced services. Many technologies and architectural patterns are proposed to develop such services. The however number of real world applications is still limited. In this paper, we present a real world deployment related to Industry 4.0. This article details our technical choices and the benefits of the approach, mostly based on service-oriented technology. First and foremost, we also point out the specific requirements and constraints we had to face in this work.

Tour recommendation is popular nowadays for providing the best-fit route plans to tourists. Existing applications only focus on the sequences design of Points of Interest (POIs) while ignore the detailed information in large-scale POIs. To further satisfy tourists’ demands, we propose Embedded Tour (eTOUR), a two-layer framework that takes these large-scale POIs, which we call Super-POIs, into account. The framework is first divided into Outer Model and Inner Model and then combined by an Embedded GRASP-VNS Algorithm based on an embedding strategy. For Outer Model, we apply Greedy Randomized Adaptive Search Procedure (GRASP) for route construction and Variable Neighborhood Search (VNS) for local search. During the outer route construction process, Super-POI is first treated as a “meta node” and in the late period, inner route is revised dynamically to adapt to the outer route where DFS-based Tree Search with Pruning is applied. Furthermore, we consider a special case in the Super-POI and modify the solution of Chinese Postman Problem to reduce the complexity. Finally, experiments based on real datasets demonstrate the effectiveness of our proposal.

Applications are increasingly containerized using techniques, such as LXC and Docker. Scientific workflow applications are no exception. In this paper, we address the problem of resource contention between concurrently running containerized scientific workflows. To this end, we design and implement Hierarchical Recursive Resource Sharing (HRRS), which structures multiple concurrent containers in a hierarchy that automatically and dynamically regulates their resource consumption based on their level/tier in the hierarchy. The hierarchy is recursively updated as the top-tier container completes its execution with the second-tier container becoming the top-tier container inheriting the resource consumption priority. We have evaluated the performance of HRRS using multiple large-scale scientific workflows containerized by Docker. The experimental results show the significant reduction of resource contention as evident in performance improvement of 49%, 160% and 18% compared with sequential execution, concurrent execution with fair resource share and execution with submission interval, respectively.

Cloud computing provided the elasticity for its users allowing them to add or remove virtual machines depending on the load of their web applications. However, there is still no ideal auto-scaler which is both easy to use and sufficiently accurate to make web applications resilient under the dynamic load. The threshold-based auto-scaling approaches are among the most popular reactive auto-scaling strategies due to their high learnability and usability. However, the static threshold would become undesirable once the workload becomes highly dynamic and unpredictable. In this paper, we propose a novel fuzzy logic based approach that automatically and adaptively adjusts thresholds and cluster size for a web application. The proposed auto-scaler aims at reducing resource consumption without violation of Service Level Agreement (SLA). The performance evaluation is conducted with the real-life Wikipedia traces in the Amazon Web Services cloud platform. Experimental results demonstrate that our reactive auto-scaler efficiently reduces cloud resources usage and minimizes the SLA violations.

Continuous deployment techniques support rapid deployment of new software versions. Usually a new version is deployed on a limited scale, its behavior is monitored and compared against the previously deployed version and either the deployment of the new version is broadened, or one reverts to the previous version. The existing monitoring approaches, however, do not capture the differences in the execution behavior between the new and the previously deployed versions.We propose an approach to automatically discover execution behavior models for the deployed and the new version using the execution logs. Differences between the two models are identified and enriched such that spurious differences, e.g., due to logging statement modifications, are mitigated. The remaining differences are visualized as cohesive diff regions within the discovered behavior model, allowing one to effectively analyze them for, e.g., anomaly detection and release decision making.To evaluate the proposed approach, we conducted case study on Nutch, an open source application, and an industrial application. We discovered the execution behavior models for the two versions of applications and identified the diff regions between them. By analyzing the regions, we detected bugs introduced in the new versions of these applications. The bugs have been reported and later fixed by the developers, thus, confirming the effectiveness of our approach.

Cloud-based computing systems could get oversubscribed due to budget constraints of cloud users which causes violation of Quality of Experience (QoE) metrics such as tasks’ deadlines. We investigate an approach to achieve robustness against uncertain task arrival and oversubscription through smart reuse of computation while similar tasks are waiting for execution. Our motivation in this study is a cloud-based video streaming engine that processes video streaming tasks in an on-demand manner. We propose a mechanism to identify various types of “mergeable” tasks and determine when it is appropriate to aggregate tasks without affecting QoS of other tasks. Experiment shows that our mechanism can improve robustness of the system and also saves the overall time of using cloud services by more than 14%.

Fairness and efficiency are generally two important metrics for users in modern cloud computing. Due to the heterogeneous resource demands of CPU and memory for users’ tasks, it cannot achieve the strict 100% fairness and the maximum efficiency at the same time. Quantitatively showing the fairness degradation/loss becomes essentially important in the design of any fairness-efficiency tradeoff scheduler. Existing fairness-efficiency schedulers (e.g., Tetris) can balance such a tradeoff elastically by relaxing fairness constraint for improved efficiency using the knob. However, their approaches are insensitive to the fairness degradation under different knobs, which makes several drawbacks. First, it cannot quantitatively tell how much relaxed fairness can be guaranteed (i.e., QoS of fairness guarantee) given a knob value. Second, it fails to meet several essential properties such as sharing incentive. To address these issues, we propose a new fairness-efficiency scheduler, QKnober, to balance the fairness and efficiency elastically and flexibly using a tunable fairness knob. QKnober is a fairness-sensitive scheduler that can maximize the system efficiency while guaranteeing the $$\theta $$ -soft fairness by modeling the whole allocation as a combination of fairness-purpose allocation and efficiency-purpose allocation. Moreover, QKnober satisfies fairness properties of sharing incentive, envy-freeness and pareto efficiency given a proper knob. We have implemented QKnober in YARN and evaluated it using real experiments. The results show that QKnober can achieve good performance and fairness.

Massively Multiplayer Online Games (MMOGs) routinely have millions of registered players and hundreds of thousands of active concurrent gamers. To guarantee quality of experience (QoE) to a highly variable number of concurrent players, MMOG infrastructure have converted nowadays into cloud computing paradigm. Many leading MMOG companies have begun to build increasing numbers of energy hungry data centers for running the MMOG services requested by the players. A main challenge for MMOG service providers is to find the best tradeoff between two contradictory aims: improving the QoE and reducing energy costs. In this paper, we propose a dynamic resource provisioning scheme for large-scale MMOG services implemented on top of cloud infrastructures which takes advantage of both virtual machine resizing and server consolidation to achieve energy efficiency and desired QoE requirements. Our experimental results indicate that, compared to an over-provisioning of infrastructural resources, our resource provisioning scheme can achieve up to 54.5% energy savings while providing the just-good-enough QoE to gamers under rapidly changing workloads.

As a method of data analysis that automates analytical model building, machine learning is becoming increasingly popular. In most machine learning algorithms, hyperparameter optimization or tuning is a necessary step. Unfortunately, the process of hyperparameter optimization is usually computationally expensive and time-consuming. Currently, machine learning is becoming a service so that cost and time should be considered when a machine learning service is provided. In this paper, we propose a scheduling approach to satisfy two contradictory targets, i.e., cost and time, when models corresponding to multiple settings of hyperparameters need to be tried. In this approach, the execution time of the model with specific settings of the hyperparameters can be predicted first. Then we generate an optimized workflow instance model, which consists of multiple parallel branches and each branch sequentially executes multiple models on a server. Based on the number partitioning algorithm, the branches are organized in such a way that they have a similar execution time and can be completed almost at the same time. Through experiments on different machine learning algorithms, it demonstrated that this approach meets the deadline and reduce the cost at the same time.

Distributed platforms are widely deployed to provide services in various trades. With the increasing scale and complexity of these distributed platforms, it is becoming more and more challenging to understand and diagnose a service request’s processing in a distributed platform, as even one simple service request may traverse numerous heterogeneous components across multiple hosts. Thus, it is highly demanded to capture the complete end-to-end execution path of service requests among all involved components accurately. This paper presents REPTrace, a generic methodology for capturing the complete request execution path (REP) in a transparent fashion. We propose principles for identifying causal relationships among events for a comprehensive list of execution scenarios, and stitch all events to generate complete request execution paths based on library/system calls tracing and network labelling. The experiments on different distributed platforms with different workloads show that REPTrace transparently captures the accurate request execution path with reasonable latency and negligible network overhead.