Object Storage
Object Storage, previously known as Datasets (EOS), is backed by a bucket in E2E Object Storage (EOS), an S3-compatible store. Use it for training data, checkpoints, and artifacts that need to grow without a fixed size, be read by several workloads at once, or be reached from outside TIR with an S3 CLI or SDK.
In the TIR portal it lives under Storage > Object Storage. Each bucket has a bucket name, which is what EOS knows it as, and an alias, its name inside TIR.
Where it is used
Buckets always mount read-only. Where the bucket appears depends on what you attach it to:
- Instances (containers) and Virtual Machines: always at
/datasets/<alias>. - Slurm clusters: at a mount path you choose when you attach the bucket, such as
/data.
Buckets are also the destination for Data Syncer transfers and the store behind fine-tuning data.
API Reference
Object Storage API Reference
Create and manage buckets, objects, versioning, locks, and replication.
/datasets/Create Object Storage/datasets/List Object Storage/datasets/{Dataset_Id}/objects/List Data Objects/datasets/{Dataset_Id}/versioning/Enable or Suspend Versioning/datasets/{Dataset_Id}/lock/Set Bucket Lock Configuration/datasets/{Dataset_Id}/presigned-urlGenerate Presigned URL/datasets/{Dataset_Id}/replication/Create Replication Rule/datasets/{Dataset_Id}/objects/empty/Empty Bucket/datasets/{Dataset_Id}/Delete Object StorageBilling & Plans
Billing & Credits
Object Storage is billed on peak usage per calendar month, plus data transfer.
Peak Usage
Charged on the highest storage reached during the month.
No Fixed Size
Capacity grows with your data; nothing to pre-allocate.
Pricing Page
Slab rates, worked examples, and a Block Storage comparison.
Object Storage Documentation
Core pages for setup and usage
Related Storage
Other TIR storage options