Cursor Uses S3 WAL to Scale Git Storage to More than 300 Pushes per Second
300 pushes per second. Cursor's new Git storage architecture separates compute from consistency—and shows how cloud-native storage is reshaping developer infrastructure.
Why it matters
Cursor's Continuity architecture demonstrates a scalable pattern for developer infrastructure: S3 as the source of truth, local NVMe as a warm cache, and linear read scaling across 100+ replicas. This is a concrete example of how agentic and high-throughput AI workloads are driving rearchitecture of foundational tools—separating replica coordination from consistency is a blueprint other platforms may adopt.
The key facts
12 to knowS3-backed write-ahead log as source of truth for Git storage
Local NVMe repositories function as warm caches
Linear read scaling with up to 100 replicas reported
300+ pushes per second in synthetic tests using S3 Express One Zone
Architecture separates replica coordination from consistency
Published September 30, 2026 on InfoQ
S3-backed write-ahead log (WAL) as source of truth for Git storage
Linear read scaling reported with up to 100 replicas
More than 300 pushes per second in synthetic tests with S3 Express One Zone
Separates replica coordination from consistency
Article source: InfoQ (September 30, 2026)
No production deployment data or real-world performance metrics disclosed
Go to the source
InfoQ AI/MLinfoq.com
Publisher excerpt: Cursor has introduced Continuity, a Git storage architecture that uses an S3 backed write ahead log as the source of truth. The design turns local NVMe repositories into warm caches and separates replica coordination from consistency. Cursor reports linear read scaling with up to 100 replicas and…