Hiroshi Wada

E³: A Multiobjective Optimization Framework for SLA-Aware Service Composition

In Service-Oriented Architecture, each application is often designed as a set of abstract services, which defines its functions. A concrete service(s) is selected at runtime for each abstract service to fulfill its function. Since different concrete services may operate at different quality of service (QoS) measures, application developers are required to select an appropriate set of concrete services that satisfies a given Service-Level Agreement (SLA) when a number of concrete services are available for each abstract service. This problem, the QoS-aware service composition problem, is known NP-hard, which takes a significant amount of time and costs to find optimal solutions (optimal combinations of concrete services) from a huge number of possible solutions. This paper proposes an optimization framework, called E3, to address the issue. By leveraging a multiobjective genetic algorithm, E3 heuristically solves the QoS-aware service composition problem in a reasonably short time. The algorithm E3 proposes can consider multiple SLAs simultaneously and produce a set of Pareto solutions, which have the equivalent quality to satisfy multiple SLAs.

CloudDB AutoAdmin: Towards a Truly Elastic Cloud-Based Data Store

In this paper, we present the design and the architecture of the CloudDB AutoAdmin system which aims to fill the existing gaps between the provided cloud database services and the requirements of the consumer applications. In particular, it focuses on facilitating the job of the cloud database consumers in implementing database applications as distributed, scalable, and elastic services with a minimum effort on the side of the application developer and a limited footprint in the application code.

Cloud API issues: an empirical study and impact

Outages to the cloud infrastructures have been widely publicized and it would be easy to conclude that application developers only need to be concerned with large scale cloud provider infrastructure outages. Unfortunately, this is not the case. In-cloud applications heavily rely on cloud infrastructure APIs (directly or indirectly through scripts and consoles) for many sporadic activities such as deployment change, scaling out/in, backup, recovery and migration. Failures and/or issues around API calls are a large source of faults that could lead to application failures, especially during sporadic activities. Infrastructure outages can also be greatly exacerbated by API-related issues.n In this paper we present an empirical study of issues in Amazon EC2 APIs. Some of the major findings around API issues include: 1) A majority (60%) of the cases of API failures are related to stuck API calls or unresponsive API calls. 2) A significant portion (12%) of the cases of API failures are about slow responsive API calls. 3) 19% of the cases of API failures are related to the output issues of API calls, including failed calls with unclear error messages, as well as missing output, wrong output, and unexpected output of API calls. 4) There are 9% cases of API failures reporting that their calls (performing some actions and expecting a state change) were pending for a certain time and then returned to the original state without informing the caller properly or the calls were reported to be successful first but failed later. We also classify the causes of API issues and discuss the impact of API issues on application architectures.

Detecting cloud provisioning errors using an annotated process model

In this paper, we demonstrate the feasibility of annotating a process model with assertions to detect errors in cloud provisioning in near real time. Our proposed workflow is: a) construct a process model of the desired provisioning activities using log data, b) use the process model to determine appropriate annotation triggers and annotate the process model with assertions, c) use the process model to monitor the deployment logs as they are generated, d) trigger the assertion checking based on process activities and log entries, and e) check the assertions to determine errors.n For a production deployment tool, Asgard, we have implemented the steps involving constructing a process model, using the model to determine appropriate annotation triggers, triggering the annotation checking based on Asgard log files, and detecting errors. Our prototype has detected errors that cross deployment tool boundaries and go undetected by Asgard; it further has detected other errors substantially more quickly than Asgard would have.

Application-Managed Database Replication on Virtualized Cloud Environments

One among several patterns that are common for applications being deployed in cloud platforms is to take an existing application designed for a conventional data center, and then port it to the cloud with minimal changes. When this is done, the application tier can easily take advantage of the elasticity and scale provided by the cloud, but the data management layer, being stateful, faces more issues. In this paper, we explore experimentally the limits to scaling for an application that itself manages database replicas each placed in a virtual machine in the cloud (exactly following the design used when the application would be deployed on an in-house cluster). We characterize important limits in the load on the master copy, the workload imposed on each slave copy when processing updates from the master, and also from the increasing staleness of replicas.

Eliciting operations requirements for applications

The DevOps community advocates communication between the operations staff and the development staff as a means of ensuring that the developers understand the issues associated with operations. This paper argues that “communication” is too vague and that there are a variety of specific and well known sources that developers can examine to determine requirements to support the installation and operations of an application product. These sources include standards, process descriptions, studies about sources of failure in configuration and upgrade, and models that include both product and process.

Availability analysis for deployment of in-cloud applications

Deploying critical applications in the cloud introduces uncertainties for availability that have traditionally been under the direct control of the application owner. The cloud infrastructure impact to availability is due to dynamic resource sharing as well as limited visibility/control of the underlying infrastructure and its quality of service. It is important to assess the availability of the critical application considering the weak availability guarantees provided by the cloud infrastructures under a broad range of scenarios, including rare scenarios like infrastructure failures and disasters. In this paper, we propose a deployment architecture-driven availability analysis model that considers uncertain rare events explicitly and bridges the gap of weak infrastructure availability and critical application availability. The models require initial calibration and validation, which is achieved by using data from commercial products and industry best practices. We use the proposed models to reevaluate the industry best practice under rare infrastructure events.

Management towards reducing cloud usage costs

Many organizations are attracted to cloud computing as an ICT (information and communications technology) sourcing model that can improve flexibility and total cost of ICT systems. However, it can be difficult for a prospective cloud customer to determine and manage cloud usage costs. We present an overview of several NICTA research projects that aim at providing information that can help ICT professionals determine various cloud usage costs and make decisions that are appropriate from the business viewpoint. Before migrating an application into a cloud, it is necessary to choose to which cloud to migrate, because there is a huge variety of cloud offerings, with significantly different pricing models. To accurately capture projected operating costs of an application in a particular cloud and enable side-by-side comparison of cloud offerings from different providers, NICTA developed a cost estimation tool that calculates the costs based on usage patterns and other characteristics of the application. This tool can also be used during runtime as an input into making adaptation/control decisions. To collect various runtime metrics (e.g., about the amount of transferred data or received quality of service - QoS) that are necessary for operational management and assessment of cloud usage costs, NICTA developed an innovative tool for flexible and integrated monitoring of applications in clouds and (in case of hybrid clouds) related local data centers. To help determine which runtime adaptation/control decisions are best from the business viewpoint (e.g., incur lowest cost), we extended the WS-Policy4MASC language and MiniZnMASC middleware for autonomic business-driven IT management with events and adaptation actions relevant for cloud management. The tools from the presented projects can be used separately or as parts of a powerful integrated cloud management system (which contains several additional tools).