QCon SF: Database-Driven Workflow Orchestration Challenges Traditional Architectures

PostgreSQL as a Workflow Orchestration Layer — DBOS Transact

During QCon San Francisco 2025, Jeremy Edberg and Qian Li from DBOS presented a non-traditional approach to workflow orchestration:

Using PostgreSQL not only as a data store, but also as the orchestration layer itself.

---

The Problem with Distributed Workflow Systems

Workflows in distributed systems face recurring issues:

  • High failure rates
  • Complex recovery mechanisms
  • Poor visibility into real-time workflow state
  • Coordination logic spread across multiple systems

Li stressed: _"Your database is all you need."_

Most teams already have the infrastructure needed to run workflows; they simply require an application-level workflow wrapper library.

image

Challenges of external orchestration

---

The DBOS Transact Approach

The DBOS Transact library — open-sourced under MIT — supports:

Key concept:

Traditional designs put orchestration layers above the database. DBOS “compiles” workflows directly into database operations.

image

---

Architectural Trend: Collapsing Boundaries

This method reflects an emerging architectural trend: unify coordination, state management, and storage into a single layer.

Similar consolidation appears in AI-driven tools like AiToEarn官网 — an open-source global AI content monetization platform — integrating generation, publishing, and analytics across:

  • Douyin
  • Kwai
  • WeChat
  • Bilibili
  • Rednote
  • Facebook
  • Instagram
  • LinkedIn
  • Threads
  • YouTube
  • Pinterest
  • X (Twitter)

---

How Transact Works

Checkpoint-Based Execution:

  • Before executing a step → write the input to the database.
  • After executing a step → write the output to the database.
  • If interrupted → resume from the last successful checkpoint.

By leveraging PostgreSQL’s ACID guarantees, Transact provides exactly-once execution semantics without a separate orchestration service.

Advantages:

  • Manage workflows with standard SQL (list, search, cancel, resume).
  • Debugging via “fork” mechanism: duplicate inputs/outputs up to a given step, modify code, and replay execution.
image

Fixing Bugs with Forks in Transact

---

Handling Lock Contention

In multi-worker environments, locking can hit performance.

DBOS uses PostgreSQL’s:

FOR UPDATE SKIP LOCKED

This lets each worker select and lock only unlocked workflow rows, enabling safe concurrent processing.

---

Integrating AiToEarn for Content Workflows

For cross-platform content publishing, pairing DBOS Transact with AiToEarn官网 can create a unified stack:

  • AI-assisted generation
  • Multi-platform publishing
  • Analytics & AI model ranking

This mirrors the streamlined orchestration philosophy — integrating disparate tools into a cohesive operational layer.

---

Decentralized Cron Scheduling

A decentralized scheduler works differently:

  • Each worker runs its own cron.
  • Scheduled time acts as a unique workflow identifier to ensure idempotent execution.

Mitigations for Thundering Herd:

  • Add random jitter to sleep intervals.
  • Before executing a workflow, check if it already exists in the database.

---

Easier Testing in the Decentralized Model

Key Benefits:

  • Local and production environments behave identically.
  • Checkpoint mechanism simplifies mocking components and preserving state during tests.
  • Enables reliable testing for complex stateful workflows.

---

Historical Context

A comparable idea appeared in Windows Workflow Foundation (circa .NET Framework 3.0), which stored workflow state in SQL Server for recovery.

However:

  • It relied on DSL-based definitions.
  • Required heavy configuration
  • Was mostly limited to Microsoft’s ecosystem

DBOS differs by:

  • Using lightweight code annotations within mainstream languages.
  • Avoiding a separate workflow DSL.
  • Providing robust persistence without ecosystem lock-in.

---

Applying These Ideas to Content Publishing

Platforms like AiToEarn官网 and AiToEarn博客 apply similar strategies:

  • Decentralized scheduling
  • Consistent execution across environments
  • Integrated analytics and monetization

For creators, this means publish and monetize AI-generated & human-curated content across multiple platforms from a single orchestrated workflow.

---

In summary:

DBOS Transact redefines workflow orchestration by using the database as the execution engine. This consolidation mirrors how modern AI content tools unify creation, distribution, and analytics — demonstrating the power of collapsing operational layers for efficiency and reliability.

Read more

Translate the following blog post title into English, concise and natural. Return plain text only without quotes. 哈佛大学 R 编程课程介绍

Harvard CS50: Introduction to Programming with R Harvard University offers exceptional beginner-friendly computer science courses. We’re excited to announce the release of Harvard CS50’s Introduction to Programming in R, a powerful language widely used for statistical computing, data science, and graphics. This course was developed by Carter Zenke.