State of the Art on Microservices Autoscaling: An Overview

Abstract

The adoption of microservices architecture has taken on great pro-portions due to its benefits and popularization of containers driven tools, such as Kubernetes and Docker. Besides, the development of microservice-based applications is a complex task, specially because they can be composed of multiple heterogeneous parts. In particular, one of the main challenges is how to conduct the microservices autoscaling (i.e., adding or removing resources on demand), while still avoiding resource waste, such as CPU and memory. This paper presents the state of the art of approaches to solve the problem of micro services autoscaling, the main characteristics to be considered as well as the important future directions that need to be still investigated.

A review of microservices autoscaling with formal verification perspective

“…In this section, we describe our systematic literature review (SLR) based on the process outlined in Pahl and Jamshidi, 13 Nunes et al 6 and Bushong et al 5 to study the existing research efforts that contributed to three research topics which are microservice, verification, and cloud autoscaling. As a result, our literature review process follows three main phases which are planning , conducting , and reporting .…”

“…The last challenge is encouraging fast scaling action of microservices. The reactive autoscaling approach may not be enough in dealing with the workloads received by the microservice systems as it can lead to problems such as cold initiation of replicas and delay in resource addition process 6 . On the other hand, the proactive autoscaling approach which considers load forecasting may be involved with a long training phase before providing the optimal decisions 22 …”

“…In horizontal scaling approach, the service instances or replicas is added to or removed from a microservice system whilst for vertical scaling, the capacities of its computing resources (i.e., CPU and RAM) will either be increased or decreased. In addition, horizontal scaling is needed to meet the demands if the number of service requests increases or decreases whilst vertical scaling is required to meet the needs of services that process tasks that require a lot of resources 6 . Meanwhile, the decision made using the two scaling dimension might be focusing on several factors such as discussed in Section 3.2.4.…”

“…The threshold values for autoscaling are usually related to resource utilization (i.e., CPU, memory) and a maximum number of the service instances to be replicated 32 . Although this strategy may not lead to excessive resource provisioning, several drawbacks of the implementation have been acknowledged which involve cold initiation of service instances, causes a delay in resource allocation as well a lack of information on the monitored services 6 …”

“…The proactive strategy is expected to be able to improve microservice autoscaling in providing sufficient resources to meet future demands. However, several challenges need to be addressed such as the inability to identify the appropriate conditions to trigger scaling actions due to lack of training, the requirement for better monitoring metrics as well as the need for the historical data structure of the services 6 . Lastly, there are also hybrid scaling strategies implemented in a cloud environment that combine the two approaches 35 and several existing research that implement this strategy in microservice autoscaling address the SLA‐conflict 36 and QoS (i.e., response time) factors 32 …”