Community Edition
Self Hosted
Cloud Service
Cloud Hosted
Everything in Community Edition, plus:
Enterprise Edition
On-premises (starting from 100TB)
Everything in Cloud Service, plus:
| Community Edition | Cloud Service | Enterprise Edition | ||
|---|---|---|---|---|
| Metadata Engine | Redis, TiKV, MySQL, PostgreSQL, etc. | High performance distributed metadata engine developed by Juicedata | High performance distributed metadata engine developed by Juicedata with horizontal scalability | |
| Data Storage | 30+ Object Storage | 30+ Object Storage | 30+ Object Storage | |
| File Amount | 100 million - 10 billion | 100 million | 10 billion | |
| Data Volume | 1PB - 100PB | 10PB | 100PB | |
| Metadata Throughput | 10,000 - 1,000,000 IOPS | 100,000 IOPS | Scales horizontally to 10,000,000 IOPS | |
| Number of Clients | 100 - 10,000 | 1,000 | 10,000 | |
| POSIX | ||||
| Kubernetes CSI Driver | ||||
| Hadoop SDK | ||||
| S3 Gateway | ||||
| Local Metadata Cache | ||||
| Local Data Cache | ||||
| Metadata Cache Expiration by TTL | ||||
| Active Metadata Cache Invalidation | ||||
| Distributed Data Cache | ||||
| Independent Cache Cluster | ||||
| Global Quota | ||||
| Data Compression | ||||
| Subdirectory Quota | ||||
| Cross-cloud and Cross-region Replication | ||||
| Web Console | ||||
| Data Encryption | ||||
| Trash | ||||
| Snapshot | ||||
| POSIX ACL | ||||
| Access Control | ||||
| Kerberos | ||||
| Apache Ranger | ||||
| Support | Community support: GitHub, Slack channel | Professional team online support | Customizable support | |
| Deployment | Self-hosted | Fully managed cloud service | On-premises |
Thanks to JuiceFS, which supports concurrent access and is fully compatible with POSIX API, we are able to implement AI distributed learning and large-scale service applications on our AI platform. With JuiceFS, our system processes massive amounts of data with stability and efficiency.
NAVER Engineering Center
We built a general-purpose storage solution using JuiceFS, tailored to meet the specific needs of AI: high performance, reliability, low cost, and multi-cloud compatibility. At the same time, it can be mounted on GPU instances just like a regular filesystem. This allows us to focus on what we do best—building a cloud-native AI platform and optimizing high-performance GPU compute—while leaving storage optimization to the experts.
Lepton AI CEO
JuiceFS not only provides a storage system that meets our hybrid cloud needs, but also optimizes the data processing workflow with high-performance metadata services and multi-interface compatibility, significantly reducing operational costs. It is the ideal choice for running our LLM platform.
MiniMax Technical Director
