On-demand computing platform at Booking.com.

By vitaly.pronto

Elevator Pitch

How do we run 10M tasks per day at Booking.com or An Approach to design a distributed, fault-tolerant, horizontally scalable tasks scheduler On-demand computing platform for all Booking.com services.

Description

We created inspectable and monitored automatic distribution and scheduling system with multiple failovers from scratch, and I want to talk about issues and features. Or How do we run 10M tasks in a day

Index

  • On-demand computing benefits:
    • Use/cases
    • Simplicity
    • Utilization
  • Architecture:
    • Requirements
    • Issues during creating:
      • Load balancing
      • Scaling
      • Resilience
      • Framework deployment
    • Metrics or SLO/SLI
  • On-demand computing and Season load changes

Notes

Platform provides:

  • Resiliency
  • Reliability
  • Monitoring
  • Predictability
  • Flexibility
  • Isolation