Recently the RAINBOW Fog-aware Kubernetes scheduler has been officially contributed to and forked by the Polaris SLO Cloud project, which is one of the Special Interest Groups (SIGs) of Linux Foundation’s Centaurus project, “an open-source platform for building a unified and scalable distributed cloud infrastructure”. Their goal is to allow combining Cloud and Edge resources at a large scale and expose them as a unified infrastructure. The Polaris SIG occupies itself with making Service Level Objectives (SLOs) “first class entities in cloud computing”. [1] [2]
Polaris Scheduler
With the RAINBOW Scheduler forked under the name Polaris Scheduler, the Polaris SIG is moving towards realizing its vision of an SLO-aware scheduler for the Edge [3]. In its current state the scheduler is already capable of handling the heterogeneity of Fog/Edge resources and network SLOs.
The Service Graph is used to model the topology and specify the network SLOs of the applications that should be deployed. While in RAINBOW, the Service Graph is the main entity that is deployed and all pods are created from the Service Graph’s nodes, the Polaris Scheduler relies on a lightweight version of the Service Graph, which is used as a complement to Kubernetes native deployments. Developers rely on the Service Graph to model the relationships between their applications’ microservices and to specify the network Quality of Service requirements, i.e., SLOs. Kubernetes native deployments then need to reference the Service Graph object and a node within it using labels in their metadata, thus allowing the scheduler to discover the Service Graph and enforce its SLOs during the scheduling process.
Future Collaboration
Once Polaris Scheduler is fully integrated into Centaurus, certain SLO-aware enhancements may prove beneficial for RAINBOW as well and be backported to the upstream project. Vice-versa, improvements made on the RAINBOW Scheduler will continue to be contributed to the Polaris Scheduler and make an impact on the community of the Centaurus project and, thus, possibly future public and/or private Cloud-Edge infrastructures.
References
[1] T. Pusztai, A. Morichetta, V. C. Pujol, S. Dustdar, S. Nastic, X. Ding, D. Vij and Y. Xiong, “SLO Script: A Novel Language for Implementing Complex Cloud-Native Elasticity-Driven SLOs,” 2021, p. 21–31.
[2] T. Pusztai, A. Morichetta, V. C. Pujol, S. Dustdar, S. Nastic, X. Ding, D. Vij and Y. Xiong, “A Novel Middleware for Efficiently Implementing Complex Cloud-Native SLOs,” 2021, p. 410–420.
[3] S. Nastic, T. Pusztai, A. Morichetta, V. C. Pujol, S. Dustdar, D. Vii and Y. Xiong, “Polaris Scheduler: Edge Sensitive and SLO Aware Workload Scheduling in Cloud-Edge-IoT Clusters,” 2021, p. 206–216.
[4] T. Pusztai, F. Rossi and S. Dustdar, “Pogonip: Scheduling Asynchronous Applications on the Edge,” 2021, p. 660–670.
[5] S. Nastic, A. Morichetta, T. Pusztai, S. Dustdar, X. Ding, D. Vij and Y. Xiong, “SLOC: Service Level Objectives for Next Generation Cloud Computing,” IEEE Internet Computing, vol. 24, no. 3, p. 39–50, 2020.
