Self-managed services using MirageOS unikernels
Al Amjad Tawfiq Isstaif · 2020
Next generation service-based systems will become increasingly finer-grain, larger-scale with heterogeneous performance characteristics. These services will need to be orchestrated across multiple IT infrastructures with application-specific auto-scaling capabilities to self-adapt to workload and environment changes. Today's generic autoscalers fail to address the adaptation needs of such systems and current methods for developing effectively application-aware autoscalers require substantial efforts and are far from practical. In this PhD project, I propose that distributed tracing alongside library operating systems (unikernels) can provide a cost-effective alternative approach to build self-scaling services. Such services will embody their own auto-scaling systems that are both generic and application-aware.