Santiago Gómez Sáez

Optimal Distribution of Applications in the Cloud

The emergence of the cloud computing paradigm introduces a number of challenges and opportunities to application and system developers. The multiplication and proliferation of available offerings by cloud service providers, for example, makes the selection of an appropriate solution complex and inefficient. On the other hand, this availability of offerings creates additional possibilities in the way that applications can be engineered or re-engineered to take advantage of e.g. the elastic nature, or the pay per use model of cloud computing. This work proposes a formal framework which allows to explore the possibility space of optimally distributing application components across cloud offerings in an efficient and flexible manner. The proposed approach introduces a set of innovative in their use concepts and demonstrates how this framework can be used in practice by means of a running scenario.

Enabling tenant-aware administration and management for JBI environments

Enterprise Service Buses (ESBs) constitute a core middleware technology for each modern Service-Oriented Architecture (SOA) solution. Given the popularity of the Cloud paradigm, which is based on fundamental SOA concepts, it is only therefore natural to look into how ESBs can be transformed into native building blocks for Cloud platforms. As a first step of this effort, in this work we investigate how ESBs can become multi-tenant aware, i.e. able to support multiple tenants and their users sharing the same ESB instance. A generalized architecture based on the JBI specification implemented by a number of open source ESBs is presented for this purpose. We demonstrate the feasibility of our proposal by means of a proof of concept realization and we evaluate the performance of our solution against a non multi-tenant ESB that was used as the baseline for our implementation.

Towards modeling and execution of Collective Adaptive Systems

Collective Adaptive Systems comprise large numbers of heterogeneous entities that can join and leave the system at any time depending on their own objectives. In the scope of pervasive computing, both physical and virtual entities may exist, e.g., buses and their passengers using mobile devices, as well as city-wide traffic coordination systems. In this paper we introduce a novel conceptual framework that enables Collective Adaptive Systems based on well-founded and widely accepted paradigms and technologies like service orientation, distributed systems, context-aware computing and adaptation of composite systems. Toward achieving this goal, we also present an architecture that underpins the envisioned framework, discuss the current state of our implementation effort, and we outline the open issues and challenges in the field.

A GENTL Approach for Cloud Application Topologies

The availability of an increasing number of cloud offerings allows for innovative solutions in designing applications for the cloud and in adapting existing ones for this environment. An important ingredient in identifying the optimal distribution of an application in the cloud, potentially across offerings and providers, is a robust topology model that can be used for the automated deployment and management of the application. In order to support this process, in this work we present an application topology language aimed for cloud applications that is generic enough to allow the mapping from other existing languages and comes with a powerful annotation mechanism already built-in. We discuss its supporting environment that we developed and show how it can be used in practice to assist application designers.

Performance and Cost Evaluation for the Migration of a Scientific Workflow Infrastructure to the Cloud

The success of the Cloud computing paradigm, together with the increase of Cloud providers and optimized Infrastructure-as-a-Service (IaaS) offerings have contributed to a raise in the number of research and industry communities that are strong supporters of migrating and running their applications in the Cloud. Focusing on eScience simulation-based applications, scientific workflows have been widely adopted in the last years, and the scientific workflow management systems have become strong candidates for being migrated to the Cloud. In this research work we aim at empirically evaluating multiple Cloud providers and their corresponding optimized and nonoptimized IaaS offerings with respect to their offered performance, and its impact on the incurred monetary costs when migrating and executing a workflow-based simulation environment. The experiments show significant performance improvements and reduced monetary costs when executing the simulation environment in off-premise Clouds.

Evaluating Caching Strategies for Cloud Data Access Using an Enterprise Service Bus

Nowadays different Cloud services enable enterprises to migrate applications to the Cloud. An application can be partially migrated by replacing some of its components with Cloud services, or by migrating one or multiple of its layers to the Cloud. As a result, accessing application data stored off-premise requires mechanisms to mitigate the negative impact on Quality of Service (QoS), e.g. due to network latency. In this work, we propose and realize an approach for transparently accessing data migrated to the Cloud using a multi-tenant open source Enterprise Service Bus (ESB) as the basis. Furthermore, we enhance the ESB with QoS awareness by integrating it with an open source caching solution. For evaluation purposes we generate a representative application workload using data from the TPC-H benchmark. Based on this workload, we then evaluate the optimal caching strategy among multiple eviction algorithms when accessing relational databases located at different Cloud providers.

Cloud Adaptation and Application (Re-)Distribution: Bridging the Two Perspectives

Cloud developers have to make several decisions when running their application in a cloud environment that may lead to conflicting objectives, inefficient deployment, and inappropriate or not existing adaptation strategies. Proper decision-support tools and processes are therefore needed to make cloud developers aware of the issues that need to be considered when deploying and running applications in the Cloud. Current decision support tools for cloud developers do not provide a structured and organized process in which the cloud developers can systematically check their choices when planning the deployment, execution, and adaptation of applications in the Cloud. In this paper, we combine two previous works and introduce an approach for identifying the options for (re-)deploying application in cloud providers infrastructures and the possible strategies of adaptation that can be used by the deployed application at runtime. The key contribution is a support process that synthesizes the two approaches. We also describe a case study where our support process is applied and we indicate the alternatives for application (re-)deployment and adaptation.

Implementation and Evaluation of a Multi-tenant Open-Source ESB

Offering applications as a service in the Cloud builds on the notion of application multi-tenancy. Multi-tenancy, the sharing of application instances and their underlying resources between users from different organizational domains, allows service providers to maximize resource utilization and reduce servicing costs per user. Realizing application multi-tenancy however requires suitable enabling mechanisms offered by their supporting middleware. Furthermore, the middleware itself can be multi-tenant in a similar fashion. In this work we focus on enabling multi-tenancy for one of the most important components in service-oriented middleware, the Enterprise Service Bus (ESB). In particular, we discuss the prototype realization of a multi-tenant aware ESB, using an open source solution as the basis. We then evaluate the performance of our proposed solution by an ESB-specific benchmark that we extended for multi-tenancy purposes.

Dynamic Tailoring and Cloud-Based Deployment of Containerized Service Middleware

The emergence and consolidation of container-based virtualization techniques has simplified and accelerated the development, provisioning, and deployment of applications for the Cloud. When considering the case of composite service-based applications that rely on service middleware solutions for their operation, container-based virtualization offers the opportunity for rapid and efficient building and deployment of lightweight, optimally configured middleware instances. As such, it provides an ideal tool for the purposes of cloudifying existing middleware solutions and offering them as part of larger PaaS offerings. As part of this effort, our investigation focuses on leveraging and evaluating a container-based virtualization environment towards enabling the assembly, provisioning, and execution of dynamically tailored instances to satisfy service middleware communication requirements of specific applications. For these purposes we scope the discussion on one particular type of messaging middleware for composite service applications, the Enterprise Service Bus (ESB) technology.

Towards Dynamic Application Distribution Support for Performance Optimization in the Cloud

The Cloud computing paradigm emerged by establishing new resources provisioning and consumption models. Together with the improvement of resource management techniques, these models have contributed to an increase in the number of application developers that are strong supporters of partially or completely migrating their application to a highly scalable and pay-per-use infrastructure. In this paper we derive a set of functional and non-functional requirements and propose a process-based approach to support the optimal distribution of an application in the Cloud in order to handle fluctuating over time workloads. Using the TPC-H workload as the basis, and by means of empirical workload analysis and characterization, we evaluate the application persistence layers performance under different deployment scenarios using generated workloads with particular behavior characteristics.