---
updatedAt: 2026-06-07T14:58:34.000Z
agentTools:
siteIndex: https://docs.jfrog.com/llms.txt
projectIndex: https://docs.jfrog.com/installation/llms.txt
---
# Filestore Selection
Choose object storage (S3/GCS/Azure), NFS, or local filesystem for Artifactory artifact binaries based on topology needs.
Choose the **filestore** backend for Artifactory artifact binaries; the right choice affects performance, scalability, and HA.
## When To Use This Page
Use this page when defining binary storage before production deployment or HA rollout.
```mermaid
flowchart TD
subgraph flow["Filestore Selection"]
A[Decision Maker] -->|FileStore Selection| B{Deployment type?}
B -->|Single Node / Eval| C[Local filesystem or PVC
Simplest setup]
B -->|Production / HA| D{Cloud Provider?}
D -->|Yes| E[NFS or cluster-file-system]
D -->|On-premises| F{Shared storage
available?}
F -->|Yes| E
F -->|No| G[Consider object storage
e.g. MinIO]
E -->|AWS| H[Amazon S3
Recommended]
E -->|GCP| I[Google Cloud Storage
Recommended]
E -->|Azure| J[Azure Blob Storage
Recommended]
end
```
## Recommendation
Use Object Storage For Production
Use object storage (S3, GCS, or Azure Blob) for production deployments. Object storage provides unlimited scalability, built-in durability, and is required for HA configurations where multiple Artifactory nodes share the same binary store.
## Quick Decision
* Choose **object storage** for all production and HA deployments.
* Use **NFS** only when object storage is not available and your NFS performance is proven.
* Use **local filesystem** only for single-node evaluation and development.
## Comparison
| | **Object Storage** (S3/GCS/Azure Blob) | **NFS / Shared Filesystem** | **Local Filesystem** (PVC) |
| :------------------------------- | :------------------------------------- | :--------------------------------------- | :---------------------------- |
| **Best for** | Production, HA, cloud | Self-managed HA without object storage | Evaluation, single node |
| **HA support** | Yes — all nodes share the same bucket | Yes — all nodes mount the same NFS share | No — PVC is bound to one node |
| **Scalability** | Unlimited | Limited by NFS capacity | Limited by disk size |
| **Durability** | 99.999999999% (11 nines, S3) | Depends on NFS setup | Depends on disk/volume |
| **Performance** | High (supports direct upload/download) | Medium (depends on NFS latency) | High (local disk) |
| **Cost** | Pay per GB stored + requests | NFS infrastructure cost | Disk cost |
| **Direct cloud upload/download** | Yes (reduces Artifactory CPU load) | No | No |
| **Maintenance** | Managed by cloud provider | Requires NFS administration | Minimal |
## When To Use Each Option
### Object Storage (Production Recommended)
Use object storage when all of the following are true:
* You are deploying on **AWS, GCP, or Azure** (native integration)
* You need **HA** with multiple Artifactory nodes
* You want **unlimited scalability** without managing disk capacity
* You want to leverage **direct cloud upload/download** for better performance
| Cloud Provider | Service | Direct Upload |
| :------------- | :------------------- | :------------------------------------------------------------------------------------------------------------- |
| **AWS** | Amazon S3 | Yes ([S3 direct upload](s3-binary-storage-templates)) |
| **GCP** | Google Cloud Storage | Yes ([GCS direct upload](google-storage-binary-provider-native-client-template)) |
| **Azure** | Azure Blob Storage | Yes ([Azure direct upload](azure-blob-storage-v2-binary-provider#azure-direct-upload-v2-template-recommended)) |
Keep Bucket And Cluster In One Region
The object storage bucket **must** be in the same region as your Artifactory deployment for acceptable latency and to avoid cross-region data transfer costs.
### NFS / Shared Filesystem
Use NFS when all of the following are true:
* You are deploying in a **self-managed environment** without access to object storage
* You need **HA** but cannot use S3/GCS/Azure Blob
* Your NFS infrastructure provides **sub-5ms latency** and **high throughput**
Requirements:
* Type: SSD-backed
* Latency: Sub-millisecond to 5ms
* IOPS: 6000+
* Throughput: 1000 MB/sec+
NFS Performance Warning
NFS is **not recommended for high-load environments** due to latency sensitivity. If your NFS has inconsistent performance, artifacts may fail during upload or download.
### Local Filesystem (Evaluation Only)
Use local filesystem when all of the following are true:
* You are running a **single-node evaluation** or development environment
* You do not need HA
* You are using the **bundled database** and default settings
Local Filesystem Is Not HA-Safe
**Do not use local filesystem for HA.** A local PVC is bound to a single Kubernetes node. Multiple Artifactory pods cannot share it.
## Filestore Sharding
For additional durability and availability, Artifactory supports **filestore sharding** — distributing binaries across multiple storage volumes.
See [Configure a Sharding Binary Provider](configure-a-sharding-binary-provider) for setup instructions.
## Local Cache (All Deployments)
Regardless of the filestore backend, Artifactory uses a **local cache** (`cache-fs`) on each node's disk for frequently accessed artifacts. This significantly improves download performance.
| Spec | Recommendation |
| :--------- | :------------- |
| Type | SSD |
| Size | 500 GB+ |
| IOPS | 6000+ |
| Throughput | 1000 MB/sec+ |
Larger cache sizes improve download performance for hot artifacts. See [Sizing](hardware-sizing-matrix) for per-template recommendations.
For detailed filestore configuration, see [Filestore Configuration](filestore-configuration-landing).
## Related Topics
* [Database Selection](database-selection)
* [Installation Method Selection](installation-method-selection)
* [Single Node vs High Availability](single-node-vs-high-availability)
* [Filestore Configuration](filestore-configuration-landing)
* [Requirements Matrix](requirements-matrix)
* [Decision Trees](decision-trees)