FREEDA is an Italian research project aimed at supporting DevOps engineers in achieving failure-resilient and environmentally sustainable deployments of microservice-based applications over the Cloud-edge computing continuum. In this article, we describe motivations, objectives and first results of the FREEDA project, and discuss how FREEDA relates to the Information Systems Engineering community.
The extensive integration of smart connected devices, coupled with the increasing
computational capacities they ofer, requires a transformation of Cloud computing into ubiquitously
distributed infrastructures that exploit computing capabilities at the edge of the network
(CloudIoT) [
In this context, the widespread adoption of Microservice-based Applications (MSAs) in delivering enterprise solutions has increased the need for facilitating MSA deployment across the Cloud-IoT infrastructures seamlessly. An MSA is an application designed as a set of loosely coupled smaller components, each one with specific functionalities. Each component is also characterised by specific functional (e.g., amount of computational and storage resources needed) and non-functional (e.g., response time, latency) requirements that need to be considered and satisfied when the application is deployed on the infrastructure.
Given the complexity of both the Cloud-IoT infrastructure (i.e., number of nodes,
heterogeneity, and variability) and the application (e.g., number of components and specific requirements),
properly mapping each component to a feasible infrastructural node is becoming a complex and
time-consuming task. This calls for approaches where the diferent deployment requirements of
the microservices composing the application are considered together with the capabilities and
features of the infrastructure. These approaches could influence the DevOps practice, which
must take into consideration potential service and hosting node failures, as well as cascading
failures where the malfunction of one infrastructure node or service leads to the failure of
others [
The need for the deployment of resilient applications must also align with the recent European
Union’s requirements for sustainable IT [
The combination of all these characteristics makes it dificult to reason on the Cloud-IoT deployment configuration of MSAs. Moreover, deployment requirements can conflict (e.g., resilience can be increased by deploying replicated service instances, which increases the toll on the energy budget). This calls for novel techniques and tools that help DevOps engineers to reason on deployment requirements.
H I A R
Failure Analyzer Energy
Analyzer Enrichment
A+ I R+
Reasoner a+
I
R+ Multi-Criteria
Solver Trade-Off
Cloud-IoT continuum D E
DevOps Engineer
FREEDA aims to address the demand for DevOps support in deploying MSAs within Cloud-IoT infrastructures, integrating considerations for failure resilience and environmental sustainability. An overview of the proposed approach is illustrated in Fig. 1.
The project enables the holistic deployment of an MSA over a Cloud-IoT infrastructure by trading among its (possibly conflicting) deployment requirements. More precisely, FREEDA will enable analyzing an MSA and a Cloud-IoT infrastructure, together with information on the current and former MSA deployments (if any), to elicit additional requirements enabling the enforcement of the MSA deployment’s failure resilience and energy sustainability.
As shown in Fig. 1, the proposed solution is divided into two main phases (Enrichment and Trade-Of ) each one supported by two main components. The main stakeholder of the proposed approach is the DevOps Engineer willing to deploy an MSA in the infrastructure. The DevOps Engineer must provide the following information: • A description of the application components (A): the set of microservices composing the application and their mutual relationships. • A set of requirements (R): functional (computational and storage resources needed) and non-functional (performance, security, energy) requirements associated with each component. • A description of the infrastructure (I): the set of nodes in which the application can be deployed and their capabilities and cost. These nodes might belong to the DevOps organization or an external organization (e.g., a public or private cloud provider).
Additionally, to these inputs provided by the DevOps engineer, we expect to have also information coming from the historical monitoring of the infrastructure and the application, if already deployed (H).
Using these inputs, FREEDA will provide the DevOps Engineer with a valid deployment able to fulfil all the expressed constraints. The core of the methodology is the Multi-Criteria Solver component. This component formally models the deployment problem, by encoding it in a constraint modelling language and builds a solver providing the best deployment plan (D).
Before being able to perform this step, the Enrichment phase needs to be executed. This phase aims at improving the application and requirements description to ensure the fulfilment of the failure resiliency and environmental sustainability requirements. The enrichment phase consists of two main components: • Failure Analyzer: the focus of this component is to enrich the application and its requirements description with additional features. The failure identifies possible causes of failure in the application execution and includes proper sidecar integration components (e.g., circuit breakers) in A to avoid failure propagation. It also generates novel hard/soft requirements in R for enforcing failure resiliency at deployment time, e.g., avoiding deploying services on nodes that are known to fail if subject to a given load, or whose failure is predicted to occur soon based on the available historical data. • Energy Analyzer: this component develops techniques to classify the profile of the nodes in a Cloud-IoT infrastructure and correspondingly generate requirements for reducing the energy consumed by MSAs deployed on such infrastructure. Also, the relation between an MSA’s components, based on the connections/functional dependencies between microservices and information on the data they exchange are considered in the analysis to provide insights about microservice co-placement to the solver.
The output of the enrichment phase is an updated and enrichment description of the application components (A+) and requirements (R+), including failure-resiliency and energy-aware considerations. The enrichment phase reduces the design efort required from the DevOps Engineer by ensuring the satisfaction of these non-functional aspects in the deployment of the application without the need for her direct intervention.
As discussed in Sect. 1.1, FREEDA needs to operate in a dynamic environment that can afect both infrastructure and the application composition, as well as the applications’ requirements. Anytime a significant change is detected or requirements get violated, a new execution of the Multi-Criteria Solver is needed to generate a new deployment plan. To avoid the inefifciency that might result from a full re-deployment of an already running application, the Reasoner component is in charge of determining a —possibly small— subset a+ of the application components that need to be re-deployed due to violation of the requirements in R+. This continuous reasoning approach might generate a sub-optimal deployment plan while reducing the disruption of the re-deployment and ensuring the satisfaction of all the requirements.
The last relevant feature distinguishing the FREEDA approach from the state of the art is Explainability. FREEDA aims at providing DevOps engineers with proper and human-friendly explanations (E) on why/how specific deployment choices have been taken by the solver. Such explanations can also be exploited by the DevOps engineers to improve the MSA description or requirements. For instance, the explanation may suggest the deployment of additional components (e.g., circuit breakers) to reduce the risk of failure of another component, thus increasing the cost and energy consumption of the proposed solution. Based on this, a DevOps engineer may decide to refactor her service, natively including resiliency in the design.
(1) Holistic MSA Deployment over the Cloud-Edge continuum: FREEDA will develop novel
techniques to determine a suitable trade-of among the MSA’s deployment requirements,
therein including cost, hardware, software, security, failure resilience, and sustainability
requirements.
(2) Continuous Reasoning for Adaptive MSA Deployment over the Cloud-Edge continuum:
FREEDA will develop continuous reasoning-like [
FREEDA will deliver a set of techniques to fulfil the project’s objectives, which will be
prototyped to form a toolchain for planning MSA deployment across Cloud-IoT infrastructures. These
techniques correspond to the main components outlined in Fig. 1. Development within FREEDA
will adopt an iterative approach, beginning with basic implementations of proposed techniques
and gradually refining and extending them until project goals are achieved. Each element
of the approach will be provided as independent components, allowing DevOps Engineers to
activate and configure them according to specific needs and objectives for their applications. The
toolchain’s components will be developed with a service-oriented approach. While the specific
technologies and languages for implementation are still being examined, the aim remains to
leverage standard and open-source solutions whenever possible. As an example, the deployment
plan D and the descriptions of the application A, requirements R, and infrastructure I will be
articulated in a YAML1 file format, while the optimization problem will be formulated using
MiniZinc [
The practical application of these techniques will be demonstrated through realistic use cases,
illustrating the (re)configuration of MSA deployment over a testbed Cloud-IoT infrastructure.
Operations and monitoring of the deployment will leverage existing tools, such as those based
on TOSCA [
The prototype tools and their integration as a toolchain will be released as open-source software in public repositories, such as GitHub.
FREEDA has started its operations in September 2023. At the current stage, the partners have been involved in activities for the definition of: • A preliminary version of the mathematical formalization of the optimization problem, which included a detailed description of the application, the infrastructure, and the requirements; • A preliminary and simplified implementation of the solver using the model formalization to find a suitable deployment solution.
The project’s alignment with the themes of the International Conference on Advanced Information Systems Engineering (CAISE) underscores its relevance to the Information Systems Engineering community. Specifically, FREEDA addresses challenges that are of particular interest to this community, as delineated in the Call for Papers: • Microservices design and deployment: this topic is aligned with the main motivation of the project as expressed in objective O1; • Cloud- and edge-based IS engineering: this topic is aligned with the context in which
FREEDA operates as declared in objectives O1 and O2; • Context-aware, autonomous, and adaptive IS: this topic is aligned with the goal of the continuous reasoner component, adapting the deployment according to the context of execution and requirements violations as declared in objective O2.
Moreover, the project’s focus resonates with key themes from previous editions of the conference. For instance, the Intelligent Information Systems theme from CAISE 2021 acknowledged the heightened level of uncertainty faced by organizations and the growing imperative for Intelligent Information Systems that ofer trusted, adaptive, agile, and autonomous solutions. Similarly, the Resilient Information Systems theme from CAISE 2020 recognized the inherent complexity of information systems as they evolve, leading to susceptibility to various forms of degradation and failure. These topics are central to FREEDA’s objectives.
This work was supported by the project FREEDA (CUP: I53D23003550006), funded by the frameworks PRIN (MUR, Italy) and Next Generation EU.