❗️ OpenAI is building a massive scale storage system for its next billion users OpenAI infrastructure to hit a new bottleneck: physical space. The company is building its own massive storage system to keep ChatGPT running as usage continues to grow. Its existing platform, Habitat, already serves more than 500 Mb of data across 40 regions, supporting more than a billion users and handling more than 70 million requests per second. Azure Cosmos DB account alone processes more than 20 million queries per second, says OpenAI is unprecedented for the service. But data centers have limited space, power and cooling. OpenAI is increasingly turning to a trade-off: use this ability to store data closer to users for faster answers, or dedicate it to run more AI models. The new system combines Rockset's RocksDB cloud with Rapat-based replication to reduce flash storage requirements, sacrificing some peak query throughput through Spot in return. The terms:

❗️ OpenAI is building a massive scale storage system for its next billion users OpenAI infrastructure to hit a new bottleneck: physical space. The company is building its own massive storage system to keep ChatGPT running as usage continues to grow. Its existing platform, Habitat, already serves more than 500 Mb of data across 40 regions, supporting more than a billion users and handling more than 70 million requests per second. Azure Cosmos DB account alone processes more than 20 million queries per second, says OpenAI is unprecedented for the service. But data centers have limited space, power and cooling. OpenAI is increasingly turning to a trade-off: use this ability to store data closer to users for faster answers, or dedicate it to run more AI models. The new system combines Rockset's RocksDB cloud with Rapat-based replication to reduce flash storage requirements, sacrificing some peak query throughput through Spot in return. The terms: