Distributed Job Scheduler

Interview-Ready Approach

1) Clarify Scope and SLOs

• •Problem statement: Design a distributed cron/job scheduler that runs millions of scheduled tasks reliably with retries and monitoring.
• •Design for a peak load target around 2,894 RPS (including burst headroom).
• •Registered jobs: ~5,000,000
• •Executions per day: ~50,000,000
• •Schedule accuracy: Within 5 seconds of target time
• •Missed execution tolerance: Zero
• •Max retry attempts: 5

2) Capacity Planning Method

• •Convert traffic and growth constraints into request rate, storage growth, and concurrency budgets.
• •Keep at least 2-3x safety margin per tier (ingress, compute, storage, async workers).
• •Reserve explicit latency budgets per hop so p95 can be defended in review.

3) Architecture Decisions

• •Databases: Define a clear system-of-record and design read/write paths separately before adding optimizations.
• •Message Queues: Move non-blocking and retry-heavy work to async consumers with explicit retry and DLQ policies.
• •Monitoring: Instrument golden signals (latency, traffic, errors, saturation) per tier and per tenant/domain.
• •API Design: Standardize API boundaries, idempotency keys, pagination, and error contracts first.

4) Reliability and Failure Strategy

• •Use strong write constraints (transactions or conditional writes) and explicit backup/restore strategy.
• •Guarantee idempotent consumers and trace every message with correlation IDs.
• •Alert on user-impact SLOs, not only infrastructure metrics.
• •Apply strict input validation and backward-compatible versioning.

5) Validation Plan

• •Run one peak-load test, one dependency-degradation test, and one failover test.
• •Verify idempotency for all retried writes and async consumers.
• •Track user-facing SLOs first: p95 latency, error rate, and successful throughput.

6) Trade-offs to Call Out in Interviews

• •Databases: SQL gives stronger transactional guarantees; NoSQL often gives better write scaling and flexibility.
• •Message Queues: Async pipelines absorb spikes well, but increase eventual-consistency complexity.
• •Monitoring: Deep observability speeds incident response but raises ingestion and tooling costs.
• •API Design: Rich APIs improve developer speed but can create long-term compatibility burden.

Practical Notes

• •Partition jobs across scheduler nodes (e.g., by hash of job_id). Use a consensus protocol (Raft) or a lease-based system to reassign partitions on node failure.
• •Pre-compute the next execution time for each job and store it in a 'next_fire' column. Poll for jobs where next_fire ≤ now().
• •Use a database row-level lock or an optimistic lock to prevent two nodes from executing the same job.