RETURN TO SYSTEM DESIGN VAULT (55)BLUEPRINT #23 // Easy
#23EasyCore Distributed Systems
Design a Job Scheduler
ASKED IN INTERVIEWS AT:GoogleAmazonMicrosoftDoorDash
EXECUTIVE SUMMARY & PROBLEM STATEMENT
Design a distributed cron-like task execution engine that schedules, distributes, and executes millions of jobs reliably at exact timestamps with retries.
SCALE ESTIMATION & TRAFFIC ENVELOPE
DAILY ACTIVE USERS / TPS10 Million scheduled jobs executed per day
READ / WRITE TRAFFIC RATIOBalanced write and execution state transitions
5-YEAR STORAGE CAPACITYRelational DB metadata store with partition per month
FUNCTIONAL REQUIREMENTS
- [1]Schedule one-off and recurring jobs with cron expressions
- [2]Execute jobs across worker fleet
- [3]Retry failed tasks with exponential backoff
NON-FUNCTIONAL REQUIREMENTS
- •Job execution delay < 1s
- •Zero duplicate task executions
- •Scales to millions of scheduled jobs
CORE ARCHITECTURAL COMPONENTS
Scheduler Leader (Timing Wheel)
Distributed Lock Manager
Worker Task Queue (RabbitMQ)
Execution Worker Pool
Audit Heartbeat Log
DATA SCHEMA & PERSISTENCE
ScheduledJobs: job_id, scheduled_at, cron_expr, payload, status (PENDING, RUNNING, COMPLETED, FAILED)
DISTRIBUTED TRADE-OFFS & FAILURE MODES
- ⟡Timing Wheel algorithm allows O(1) scheduling and execution lookup, vastly outperforming scanning database tables every second.