MVCC
Multiversion Concurrency Control (MVCC) lets databases handle many users simultaneously by keeping old data versions.
Multiversion Concurrency Control, or MVCC, is a method databases use to let many people access and change data at the same time without interfering with each other.
Imagine a busy library. Instead of everyone waiting for one person to finish checking out a book, MVCC lets multiple people check out different books, or even the same book if there are multiple copies. It does this by keeping different versions of the data. When a change is made, the old version is kept for a short time. This way, other users who were looking at the old version can finish their task before seeing the new one. It's good for reading but can sometimes make writing data a bit slower because new versions need to be created.
MVCC is a concurrency control method used in database management systems to provide transaction isolation. It achieves this by assigning a unique identifier to each transaction and maintaining multiple versions of data items. When a transaction reads data, it sees a snapshot of the data as it existed at a specific point in time, typically the start of the transaction. Writes create new versions of data items rather than overwriting existing ones, allowing concurrent readers to continue using older versions. This approach generally improves read performance and concurrency but can lead to storage overhead from maintaining multiple versions and potential write contention if not managed carefully.
OpenAI's use of MVCC in PostgreSQL to manage 800 million ChatGPT users highlights the critical role of efficient database concurrency control in scaling large AI services. As AI applications grow in user base, robust database techniques like MVCC become essential for maintaining performance and reliability under heavy load.
- Scaling PostgreSQL to power 800 million ChatGPT usersOpenAI · OpenAI News ·
- OpenAI ChatGPT
- OpenAI uses MVCC in PostgreSQL to scale its database infrastructure to support hundreds of millions of ChatGPT users, demonstrating its importance for large-scale AI services.
- Locking
- Traditional locking mechanisms block other transactions when data is be
- Timestamp Ordering
- This method uses timestamps to determine the order of operat
MVCC was first described in a 1976 paper by J. S. Kung and Robert A. Morris. It has since become a standard feature in many modern relational database systems. Source ↗
While directly observing MVCC in action is difficult without database tools, you can experience its effect by using a service like ChatGPT. Try asking it a complex question, then immediately ask another unrelated question. The system's ability to quickly respond to both, without apparent delay from your first query, relies on efficient handling of concurrent requests, a task MVCC helps databases manage.
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- PromptRole-play a student struggling with math for AI tutor simulationSimulate a student's math struggles and uncertainty for AI tutor evaluation. · 5 min
- TemplateReusable workflow templates for healthcare teamsTeams can create and share templates for common workflows like documentation or prior authorizations. · 15 min
- TemplateTemplate for building an AI tool tailored to your workA structured approach to defining and building a custom AI tool for specific business needs. · 15 min
- TemplateEvals CSV prompt list for agent skillsA CSV file template for listing prompts used in evaluating agent skills. · 20 min
- HeuristicDouble coding agent task completion with docs linksProviding coding agents with a documentation link significantly increases their task completion rate. · 2 min
Questions people ask
- Does MVCC make databases faster?
- MVCC generally improves read performance and allows more users to access data simultaneously. However, it can sometimes slow down write operations because new data versions need to be created.
- What happens to old data versions in MVCC?
- Old data versions are kept for a period to allow ongoing transactions to complete. Eventually, they are cleaned up by a process called garbage collection to free up storage space.
First seen on Subvolts Jan 22, 2026. Updated Sep 15, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.