Self-Healing Cloud Services in Private Multi-Clouds

摘要

As clouds continue to dominate as the de facto means of consuming shared computing resources, cloud providers are required to ensure that their services always meet or exceed service level agreements in terms of availability, reliability, scalability, performance etc every hour, day and year. These challenging requirements are even exacerbated when it comes to multi-cloud environments. In the era of automation and continuous cost reductions, using humans to manually test and trigger corrective actions when needed in order to fulfill these service level agreements is considered cumbersome and too expensive. In order to thrive, cloud service providers need a solution which is efficient, cheaper to deploy and easy to operate. This paper introduces a distributed container based solution for testing and self-healing of cloud resources. The proposed solution is based on a containerized multi-agent architecture which leverages Elasticsearch, Logstash, Kibana (ELK) stack and uses a rule-based self-healing algorithm implemented using Drools rule engine and Spring Boot framework. Preliminary tests show that the solution meets its target objectives in terms of maintaining service level agreements, lowering deployment costs, high performance and operability.