Service Relationship Orchestration: Lessons Learned From Running Large Scale Smart City Platforms on Kubernetes

Abstract

Smart cities aim to make urban life more enjoyable and sustainable but their highly heterogeneous and distributed context creates unique operational challenges. In such an environment, multiple companies work together with government on applications and data streams spanning several management domains. Deploying these applications, each of which consists of several connected services, and maintaining an overview of application topologies remains difficult. Even though cloud modelling languages have been proposed to solve similar issues, they are not well fit for such a heterogeneous environment because they often require an "all or nothing" approach. Moreover, cloud modelling languages add an additional abstraction layer which rarely supports all features of the underlying platform and make it harder to reuse existing knowledge and tools. This research defines service relationships as the key element to modelling applications as topologies of services. We use this definition to pinpoint what is lacking in the state of the art Kubernetes orchestration tools and provide a blueprint for how relationship support can be added to any orchestrator. We present "orcon", a proof of concept orchestrator which extends the Kubernetes API to allow managing relationships between services by adding metadata to service definitions. Our evaluation shows this orchestrator enables lifecycle synchronization and configuration change propagation with an overhead of only 0.44 seconds per service.

Fog Native Architecture: Intent-Based Workflows to Take Cloud Native toward the Edge

“…Restrictions apply. example, the workflow manager contacts a K3S edge node and uses a Kubernetes relationship orchestrator [13] to deploy an undistributed composition of microservices (D). After a successful deployment, the workflow manager instructs the drone to start transmitting the video stream (E) to be processed by the workflow, i.e.…”

Cloud-Edge Orchestration for Smart Cities: A Review of Kubernetes-based Orchestration Architectures

“…Multiple industry partners, institutes, and public services must collaborate to achieve an efficient solution. This collaboration also comprises the appropriate sharing and management of network infrastructure, computational resources, and eventually data centers and their derivatives in edge computing [1]. Secondly, besides sharing and agreeing on a common data platform, further challenges must be considered as soon the organizational aspects and core infrastructure is set up.…”

“…Besides the traditional usage in business-related contexts, ICTs are used in urban areas, which is also known as the concept of Smart City (SC). In SCs, various technologies sustainably enhance urban life [1]. Many fundamental areas in cities are equipped with sensors to gather data from community services, like transportation, power plants, information systems, and crime detection [2].…”

“…Largescale container orchestration is a rather complex task that requires a sophisticated management. For this, various container orchestration tools emerged, like the popular Kubernetes (K8s) 1 . It offers high availability, scalability, and fault-tolerant management of a large number of containers.…”

“…It offers high availability, scalability, and fault-tolerant management of a large number of containers. In the context of SCs, a lot of 1 Kubernetes solutions emerged over time that are using K8s as orchestration system [1,[26][27][28][29]. So far, no comprehensive survey of already existing K8s-based architectures has been done to verify whether K8s is still worth considering for edge orchestration.…”