Shinichi Honiden

SanGA: A Self-Adaptive Network-Aware Approach to Service Composition

Service-Oriented Computing enables the composition of loosely coupled services provided with varying Quality of Service (QoS) levels. Selecting a near-optimal set of services for a composition in terms of QoS is crucial when many functionally equivalent services are available. As the number of distributed services, particularly in the cloud, is rising rapidly, the impact of the network on the QoS keeps increasing. Despite this, current approaches do not differentiate between the QoS of services themselves and the network. Therefore, the computed latency differs from the actual latency, resulting in suboptimal QoS. Thus, we propose a network-aware approach that handles the QoS of services and the QoS of the network independently. First, we build a network model in order to estimate the network latency between arbitrary services and potential users. Our selection algorithm then leverages this model to find compositions with a low latency for a given execution policy. We employ a self-adaptive genetic algorithm which balances the optimization of latency and other QoS as needed and improves the convergence speed. In our evaluation, we show that our approach works under realistic network conditions, efficiently computing compositions with much lower latency and otherwise equivalent QoS compared to current approaches.

Multi-objective Service Composition with Time- and Input-Dependent QoS

Optimizing the Quality-of-Service (QoS) levels of a service workflow is essential for the user satisfaction in Service-oriented Computing. For that purpose, QoS computation models are applied to reflect the actual QoS experienced by the user during service execution. Current QoS models ignore the possible dependencies of QoS attributes, such as the dependency on the time of the execution or on the input data supplied to the service. Apart from that, composition approaches consider only single workflows during service selection, narrowing the number of possible compositions. Thus, we introduce a novel QoS model that covers QoS dependencies and discuss how this model can be used to consider multiple workflows at the same time. Moreover, we adopt a multi-objective optimization approach to offer solutions varying in QoS such as finishing time and price, allowing the user to make fine-grained decisions.

Model-Driven-Development-Based Stepwise Software Development Process for Wireless Sensor Networks

To meet future demands for wireless sensor network (WSN) software, both experts and average software developers should be involved in WSN software development. However, WSN software development is difficult for the average software developer because data processing-related design and network-related design are tangled in the software. Here, we propose a software development process for WSN software by stepwise refinement. Our process enables stepwise refinement to separately address data processing-related and network-related concerns, reuse of well-defined designs, and implementations for network-related concerns prepared by the experts, and perform model-driven development to obtain source codes from models by model transformations. Additionally, we used case studies using actual WSN software development and user studies to evaluate how our proposed process can support actual WSN software development.

Towards Dynamic Evolution of Self-Adaptive Systems Based on Dynamic Updating of Control Loops

Self-adaptive systems, which enable runtime adaptation, are promising ways of dealing with environmental changes, including system intrusions or faults. Such software systems must modify themselves to better fit their environment. One of the main approaches to constructing such systems is to introduce multiple control loops. Software evolution is an essential activity for expanding this adaptation capability, and dynamic evolution has been envisaged as a way of systems adapting themselves at runtime. In this paper, we establish a development process to deal with dynamic evolution. We devise a goal model compiler to generate models for designing dynamic evolutions and a programming framework that supports dynamic deployment of control loops. We experimentally applied our approach to a system and discuss how our compiler and framework support dynamic evolution of self-adaptive systems.

Fast Wide Area Live Migration with a Low Overhead through Page Cache Teleportation

Live migration of virtual machines over a wide area network has many use cases such as cross-data center load balancing, low carbon virtual private clouds, and disaster recovery of IT systems. An efficient wide area live migration method is required because cross-data center connections have a narrow bandwidth. Page cache occupies a large portion of the memory of a Virtual Machine (VM) when it executes data-intensive workloads. We propose a new live migration technique, page cache teleportation, which reduces the total migration time of wide area live migration and has a low overhead. It detects the restorable page cache in the guest memory that has the same contents as the corresponding disk blocks. The restorable page cache is not transferred via the WAN but is restored from the disk image before the VM resumes. In this way, the IO performance degradation reduces after the migration. Evaluations show that page cache teleportation reduces the total migration time of wide area live migration and has a lower performance overhead than existing approaches.

A goal model elaboration for localizing changes in software evolution

Software evolution is an essential activity that adapts existing software to changes in requirements. Localizing the impact of changes is one of the most efficient strategies for successful evolution. We exploit requirements descriptions in order to extract loosely coupled components and localize changes for evolution. We define a process of elaboration for the goal model that extracts a set of control loops from the requirements descriptions as components that constitute extensible systems. We regard control loops to be independent components that prevent the impact of a change from spreading outside them. To support the elaboration, we introduce two patterns: one to extract control loops from the goal model and another to detect possible conflicts between control loops. We experimentally evaluated our approach in two types of software development and the results demonstrate that our elaboration technique helps us to analyze the impact of changes in the source code and prevent the complexity of the code from increasing.

Evaluating Impact of Live Migration on Data Center Energy Saving

Energy efficiency of cloud data centers is of great concern today and has been tackled by many researchers. Dynamic VM placement is a well-known strategy to improve energy efficiency of a data center. Virtual machines (VMs) under light load are consolidated into a small number of physical machines (PMs) to turn idle PMs into low-power states. Although live migration is essential for dynamic VM placement, former studies have not yet revealed how energy overhead of live migration has impact on energy efficiency of dynamic VM placement. To tackle this problem, we conducted integrated simulation of energy overhead of live migration and dynamic VM placement sing Sim Grid. We used three dynamic VM placement policies and two live migration mechanisms (existing pre-copy and an accelerated mechanism invented by us) to thoroughly evaluate the energy overhead. The results showed that in the worst case energy overhead of live migration occupies 5.8% of total energy consumption of a data center.

Validating ajax applications using a delay-based mutation technique

The challenge of validating Asynchronous JavaScript and XML (Ajax) applications lies in actual errors exposed in a user environment. Several studies have proposed effective and efficient testing techniques to identify executable faults. However, the applications might have faults that are not executed during testing, but might cause actual errors in a user environment. Although we have investigated static methods for finding ``potential faults that seem to cause actual errors if executed, developers need to confirm whether or not the potential faults are actually executable. Herein, we propose a mutation-based testing method implemented in a tool called JSPreventer. Even if the potential faults are not easily executable in a given environment, our method mutates the applications until they are executable using two delay-based mutation operators to manipulate the timing of the applications handling interactions. Thus, JSPreventer provides executable evidences of the not-easily-executable faults for developers, if it reveals actual errors by testing the mutated applications. We applied our method to real-world applications and found actual errors that developers could debug to improve their reliability. Therefore, JSPreventer can help developers validate reliable real-world Ajax applications.

Sample efficiency analysis of Neuroevolution algorithms on a quadruped robot

In reinforcement learning tasks with continuous state-action, parameterized policy search has been known to be a powerful method. Applying NeuroEvolution (NE) to optimizing the policy represented by artificial neural network (ANN) is a particularly active research field. In most cases, NE algorithms cost a large amount of trial-and-error (episode) to optimize policies. However, due to time and cost constraints, researchers and practitioners cannot repeat a number of episodes on physical robots. Thus, choosing an efficient NE algorithm is a key to optimize policies with limited time and cost. In this work, our goal is to help users to choose an efficient NE algorithm. We compare and analyze sample efficiency of two successful state-of-the-art NE algorithms: CMA-NeuroES and NEAT in a gait generation task of a quadruped robot. Moreover, we run both algorithms with various initial topologies in order to analyze the performance difference between each topology. From experimental results, we show CMA-NeuroES outperforms NEAT regardless of initial topologies when the limited number of episodes can be executed. Additional experiments conclude that the optimization method for connection weights in NEAT results in its inferior performance to CMA-NeuroES, while a probability-weighted averaging characteristic and self-adaptive factors make CMA-NeuroES to be advantageous.

Case studies on the development of wireless sensor network applications using multiple abstraction levels

Model-driven development (MDD) facilitates the development of wireless sensor network (WSN) applications because MDD can deal with models at the abstraction level to the concrete level. In the MDD for WSN applications, the developers model the dataflow at the abstract level and model the designs of a communication and task assignment at the concrete level. In our previous work we proposed a MDD process for WSN applications with three domain specific languages (DSLs). However, we did not sufficiently evaluate the description capabilities of our DSLs. For that reason, we report case studies on real-world applications to prove the feasibility of our approach. Our DSL can be used to describe both the application logic and design decisions concerning the quality of data, which are adopted in target applications. Additionally, we demonstrate the effectiveness of the DSLs at multiple abstraction levels.