The number of ICT services operated on large-scale systems is increasing, and faults in the target systems need to be detected for the stability of the services. When a fault occurs in the target system, the operators will analyze the monitoring data to identify a root cause and determine recovery actions. However, it is becoming inefficient to analyze them manually or detect faults with handcrafted rules. In this paper, we propose a method of generating human-readable fault reports by leveraging time-series data and textual metadata of metrics for detecting faults in a target system. It can jointly encode heterogeneous sources of monitoring data and can integrate asynchronous metrics of multiple services. To collect realistic datasets in a production environment, we conducted experiments by injecting faults into an e-commerce service implemented in a Kubernetes cluster. Our evaluation shows that the proposed method can generate fault reports appropriately and outperformed the baseline method in different scenarios.

Desktop as a Service (DaaS) has become widely used by enterprises. In 2020, the use of DaaS increased dramatically due to the demand to work remotely from home during the COVID-19 pandemic. The DaaS market is expected to continue growing rapidly [ 1 ]. The quality of experience (QoE) of a DaaS service has been one of the main factors to enhance DaaS user satisfaction. To ensure user QoE, the amount of cloud computation resources for a DaaS service must be appropriately designed. We propose an Intent-driven DaaS Management (IDM) framework to autonomously determine the cloud-resource-amount configurations for a given DaaS QoE requirement. IDM enables autonomous resource design by abstracting the knowledge about the dependency between DaaS workload, resource configuration, and performance from previous DaaS performance log data. To ensure the IDM framework's applicability to actual DaaS services, we analyzed five main challenges in applying the IDM framework to actual DaaS services: identifying the resource-design objective, quantifying DaaS QoE, addressing low log data availability, designing performance-inference models, and addressing low resource variations in the log data. We addressed these challenges through detailed designing of IDM modules. The effectiveness of the IDM framework was assessed from the aspects of DaaS performance-inference precision, DaaS resource design, and time and human-resource cost reduction.

Compression techniques for deep neural network models are becoming very important for the efficient execution of high-performance deep learning systems on edge-computing devices. The concept of model compression is also important for analyzing the generalization error of deep learning, known as the compression-based error bound. However, there is still huge gap between a practically effective compression method and its rigorous background of statistical learning theory. To resolve this issue, we develop a new theoretical framework for model compression and propose a new pruning method called {\it spectral pruning} based on this framework. We define the ``degrees of freedom'' to quantify the intrinsic dimensionality of a model by using the eigenvalue distribution of the covariance matrix across the internal nodes and show that the compression ability is essentially controlled by this quantity. Moreover, we present a sharp generalization error bound of the compressed model and characterize the bias--variance tradeoff induced by the compression procedure. We apply our method to several datasets to justify our theoretical analyses and show the superiority of the the proposed method.

To manage hybrid networks, telecommunications carriers should transform their network management architecture. We previously proposed a management architecture that has common management functions and data that are not specialized for a specific network. Using this architecture, carriers will be able to easily manage new networks to accelerate deployment. In this paper, we propose a resource-allocation architecture, which is an extension of our previous architecture. This architecture can manage the network-capacity amount and flexibly allocate it according to user demand. With this architecture, operation can be flexibly changed based on external specifications. This allows carriers to quickly release a new network service.

More businesses and organizations are choosing clouds to accommodate their workload, e.g., annalistic departments of enterprises use public/private cloud to deploy the analysis function of data, and telecommunication operators deploy Virtual Network Functions (NFVs) such as virtual Evolved Packet Core (vEPC) on clouds to provide telecommunication services. When utilizing the cloud, the cloud user is concerned about the functionality, reliability, performance, etc. of the cloud-based function, which we call the intent or the service requirements. However, to meet the intent, the user needs to tell the cloud resource orchestrator/controller what kinds of Virtual Machines (VMs) are needed and the amount of computing resources necessary to allocate to each VM, which we call resource requirements. In our preliminary work [1], we addressed this problem and proposed an Intent-Based Cloud Management (IBCM) framework to "translate" the cloud user's service requirements into resource requirements. In this work, we have focused on realizing a Resource Design Framework (RDF) for IBCM that decides the needed computing resource amount to fulfill the cloud user' intent/ service requirements. The main contributions of this work are as follows: (1) We have systematically analyzed factors that need to be taken into consideration when designing the resources. (2) On the basis of the analysis, we propose the architecture of RDF that is applicable for various cloud-based applications and scenarios. To the best of knowledge, this is the first work to address this issue. (3) We have validated the proposed RDF in an experimental cloud-based machine learning scenario. The experiment results show our RDF is able to infer the performance and operation state with over 90% precision, thus enabling resources to be precisely designed in accordance with performance requirements and operation policy. (4) In addition to precise resource design, we also proposed a service level agreement (SLA) violation prevention mechanism for RDF and verified its effectiveness.

We propose a method for reusing artifacts in the analysis phase that employs a knowledge base to realize the efficient development of operation support systems (OSSs). We have studied a method for sequence diagram reuse in the analysis phase, which is an important phase in software development. The reuse method is characterized by unifying the expressive form of artifacts using a task ontology and a domain ontology, and a similarity computation method based on a knowledge base including these ontologies. Furthermore, our evaluation of the method showed that it is effective in reducing the time needed to create the diagrams.

Density estimation plays an important and fundamental role in pattern recognition, machine learning, and statistics. In this article, we develop a parametric approach to univariate (or low-dimensional) density estimation based on semidefinite programming (SDP). Our density model is expressed as the product of a nonnegative polynomial and a base density such as normal distribution, exponential distribution, and uniform distribution. When the base density is specified, the maximum likelihood estimation of the polynomial is formulated as a variant of SDP that is solved in polynomial time with the interior point methods. Since the base density typically contains just one or two parameters, computation of the maximum likelihood estimate reduces to a one- or two-dimensional easy optimization problem with this use of SDP. Thus, the rigorous maximum likelihood estimate can be computed in our approach. Furthermore, such conditions as symmetry and unimodality of the density function can be easily handled within this framework. AIC is used to choose the best model. Through applications to several instances, we demonstrate flexibility of the model and performance of the proposed procedure. Combination with a mixture approach is also presented. The proposed approach has possible other applications beyond density estimation. This point is clarified through an application to the maximum likelihood estimation of the intensity function of a nonstationary Poisson process.