Architecture for IoT applications based on reactive microservices: A performance evaluation

Highlights

• Impact of running the application using proposed architecture in single container.

• IoT application performance varying deployment policies with limited CPU and memory.

• Power impact on gateways when running IoT applications varying deployment policies.

• Detecting overhead in terms of time execution, particularly during service failures.

• Impact of increased IoT traffic on the availability of IoT applications.

Abstract

The Internet has evolved from a network interconnecting computers to a complex ecosystem integrating devices of the most varied types, and enabling the amalgamation of the physical and virtual worlds. Integrating these heterogeneous devices fosters novel services and applications that generate value-added information and actionable knowledge for the end-user. Several challenges are involved in the design and building of IoT ecosystems, which have fostered research in the field, in search of patterns, guidelines, methods and tools that support such activities. In terms of architectural patterns, the microservice architectural style has been increasingly adopted in the development of IoT applications and services. Its adoption promotes some essential properties in IoT, such as scalability and extensibility. As there are always tradeoffs involved in every architectural decision, it is important to analyze whether the benefits brought by the use of microservices do not come at the expense of some loss or degradation of application performance. Recent research has analyzed the performance interference of microservices based on edge computing applications. However, a comprehensive assessment of the performance impact of characteristics inherent to the use of reactive microservices on IoT applications is still missing in the literature. In this paper, we present an experimental evaluation of the performance of IoT applications that make use of an architecture based on reactive microservices. The architecture was proposed by our group in a previous work and was tailored for reliable IoT applications running at the edge of the network. The experiments presented in this paper analyze the application performance based on various benchmark scenarios. In addition, we performed load and scalability testing of an IoT application that adopts the architecture components in a hybrid scenario (real devices and emulated devices). The results obtained were promising. The architecture had a good response to the increase in the workload, not presenting errors, crashes or instabilities due to the increase in IoT data traffic. Moreover, analyzing the overhead generated by the architecture components, there was no performance reduction or service unavailability. Such results point to the fact that the adoption of microservices in the construction of IoT systems can bring effective benefits without jeopardizing their performance due to eventually generated overheads.

Introduction

The Internet of Things has changed the scope of the Internet, which has become a network connecting a myriad of devices of different types. The integration of these heterogeneous devices has the potential to leverage new services and applications, produce value-added information and actionable knowledge for the end-user. However, it brought with it several practical and research challenges.

The challenges are related to the IoT intrinsic features. IoT ecosystems have a high degree of heterogeneity and dynamism. Applications based on instrumented physical things can arise and services become available in an unplanned way, taking advantage of a shared infrastructure. Such applications involve distributed software components that need to interact in an opportunistic way to achieve high-level user goals [1]. The demand for and availability of the various types of services and resources vary over time and geographically. In addition, contemporary IoT systems can be distributed across the continuum from data-producing devices to the cloud, exploiting resources located at the edge of the network. Running components on the network edge reduces delay and communication on the network backbone, but in general, edge nodes are more resource-constrained devices compared to a cloud data center and need to work collaboratively. Such features call for flexible, scalable, and lightweight solutions for the IoT. Design approaches are needed that promote scalability, fault tolerance, and the loose coupling between components. The adoption of different design patterns and architectural approaches can help meet such requirements in the construction of IoT systems and applications.

In this context, the microservice architecture [2] has several appealing features, such as high scalability, fine granularity, loose coupling, continuous development, among others, which make it an ideal candidate for building IoT systems and applications. Microservices can be described as “... an approach to developing a single application as a suite of small services, each running in its own process and communicating with lightweight mechanisms, often an HTTP resource API. These services are built around business capabilities and independently deployable by fully automated deployment machinery” [3].

According to Zimmermann [4], Microservices have the potential to overcome deficiencies in previous approaches of SOA implementation, by employing modern software engineering paradigms such as fault management, which can help improving applications reliability in general and availability in particular. In previous works [5], [6], our research group claimed that a reactive systems approach brings additional benefits in the quest to build reliable IoT systems. Reactive Systems are those which react continuously to their environment at a rate determined by such environment [7]. Therefore, our proposal was to combine the architectural principles that guide the construction of applications based on Microservices with the reactive systems principles, aiming at promoting the reliability of IoT applications. In particular, we were interested in improving reliability from an availability perspective, since we understand this is a major requirement for several IoT application domains. The proposed architecture consisted of: (i) a set of software components, and (ii) a software platform that concretizes the architecture and aids, at runtime, to meet the QoS requirements of IoT applications (focusing on availability). The components of the architecture were built to be deployed on Smart Gateways. Smart Gateways are devices, located at the edge of the network, where the application logic and middleware level services are executed.

As it is often the case in software systems, meeting one nonfunctional requirement has the potential to harm another, usually requiring either to make a trade off or adjusting the project to the specific requirements of a given application. In the case of IoT ecosystems, built to support a wide range of applications, it is not suitable to adopt an application-specific design. Therefore, it is necessary to negotiate the various non-functional requirements in order to obtain the best balance among them.

In this work, we aim to extensively analyze the performance of IoT applications built with the architecture proposed by Santana et al. [5]. The purpose of such an analysis is to verify whether the benefits brought by the use of reactive microservices do not harm the performance in terms of the consumption of computational resources, such as CPU, IO, network, and memory. In addition, considering the possible implementation in energy-constrained environments, we also analyzed the energy cost of the solution.

When working with microservices and their implementation on lightweight platforms such as containers, there are several decisions regarding their configuration and implementation that can impact the performance of the application. Containers provide application isolation and allow to perform upgrades and to escalate applications. In addition, they provide isolation and ensure consistent, portable application execution. Compared to virtual machines, containers consume fewer resources and produce lower overhead [8].

Motivated by the observations obtained in previous research [9], [10], [11], in this work we have defined a set of research questions to be answered, in order to support our analysis:

RQ1: What is the impact of running the application using all the components of the proposed architecture in a single container running on the Smart Gateway? Deploying the application using all the architecture components in a single container can be useful in scenarios where there is a limitation of computational resources.

RQ2: What is the behavior in terms of performance of an IoT application that adopts the proposed architecture with different deployment policies when there is a limit on memory and CPU usage?

RQ3: What is the impact in terms of power consumption on smart gateways when running an IoT application that adopts the proposed architecture through different deployment policies? The purpose of this question is to enquire whether energy consumption reduces or increases when deploying the IoT application through different deployment policies.

RQ4: What is the impact on the availability of IoT application that increasing the number of smart devices and, consequently, increasing the IoT traffic has on an IoT application that adopts the proposed architecture?

RQ5: What is the overhead in terms of time execution caused by components of the proposed architecture, particularly in situations where service failures occur? The purpose of this research question is to analyze whether the IoT Availability, Negotiation and Service Discovery components may include some bottleneck in services in the failures scenarios.

Therefore, the main contribution of this work is to investigate the performance of reactive microservices deployed through different deployment strategies in edge computing.

While there have been a number of research efforts in recent years that address different aspects of building a microservices based architecture, from efficient microservices implementation, architecture scalability analysis, load balancing issues, orchestration, and other system-level services, authors such as Alanezi and Mishra [9] point out that there is a lack of works reporting evaluations of a microservices-based architecture based on a real prototype implementation and including performance measurement according to various metrics. This paper contributes to filling this gap. By presenting a comprehensive performance evaluation of an IoT application based on reactive microservices running at the edge, we believe that the results of our work can effectively leverage the widespread adoption of microservices in IoT and guide system developers’ design decisions. In particular, the results of our evaluation confirmed that, from an architectural point of view, the components introduced following the microservices pattern, in order to promote the availability of the IoT system, did not degrade the performance under other performance metrics.