Navigation :

NFS at CSAIL

NFS (Network File System) provides high-performance network storage for CSAIL users. It is optimized for scenarios requiring fast local access to large datasets within the CSAIL network.

Key Benefits:

High performance optimized for fast I/O and large file transfers
Local access ideal for servers and workstations on the CSAIL network
Large storage supporting multi-terabyte datasets
Flexible tiers for scratch, persistent, and backed-up storage

Getting Started: Choose Your Path

I need temporary storage for computations

All users automatically have access to three scratch filesystems, each with a 1 TiB quota per user.

/data/scratch - For servers in Stata
/data/scratch-fast - For servers at our Holyoke data center (newer, faster)
/data/scratch-oc40 - For servers at our Holyoke data center (smaller and slower)

Scratch storage is perfect for intermediate files, logs, checkpoints, and temporary data. No backups, no snapshots. Files deleted after six months of non-access.

SCRATCH STORAGE IS FOR TEMPORARY FILES ONLY.

Files in /data/scratch, /data/scratch-fast, and /data/scratch-oc40 are automatically deleted if not accessed within six months. They may also be deleted sooner if the filesystem fills up.

Do NOT store:

Conda environments
Python package libraries or virtual environments
Software installations
Original research data
Anything you need to keep permanently

Why Python/Conda breaks in scratch: Python caches bytecode (.pyc files) but cleanup procedures delete plain-text source files (.py) when bytecode is accessed. Your Python environment then fails with cryptic errors because source files are missing.

For permanent storage, use production or archival tiers (see below).

I need permanent storage for research data

Request a production tier filesystem:

Daily backups and snapshots
High performance for parallel reads/writes
Appropriate for active research data
Contact TIG to request: Request new NFS filesystems

I have large reference datasets I don’t actively modify

Use archival tier storage:

Optimized for reliability, not performance
Weekly backups and snapshots
Best for read-only or rarely-updated data
Contact TIG to request: Request new NFS filesystems

Network Requirements

Network Requirement: NFS is only accessible from the CSAIL network (wired Ethernet or CSAILPrivate). It cannot be accessed remotely.

Security Note: NFS has no built-in encryption or security features. See NFS Security for important information.

NFS is only supported on CSAIL Ubuntu clients. The /data filesystem is synthesized by the NFS automounter based on configuration files installed by CSAIL Ubuntu.

Mailing List

If you use NFS, join the nfs-users mailing list for important announcements regarding NFS server operations and maintenance windows.

Essential Information

Understanding NFS

What are the different storage tiers? - Scratch, production, and archival explained
NFS rules and recommendations - Best practices for using NFS effectively
Performance recommendations - How to optimize your workload for NFS
Where is my NFS data? - Check which data center your data is in

Protecting Your Data

Snapshots and recovery - Point-in-time recovery from snapshots
Backup policy - Daily and weekly backups for production/archival tiers
Security considerations - NFS security model and best practices

Getting Help

Frequently Asked Questions - Common questions about NFS access, quotas, and troubleshooting
Request new NFS filesystems - Provision new production or archival storage
Distributing datasets from NFS - Publishing NFS data on the web
Accessing NFS from CSAIL Linux - NFS setup and configuration

Quick Reference

Check which server your data is on:

df /data/mydata

Access scratch storage locations:

/data/scratch        (Stata data center)
/data/scratch-fast   (Holyoke data center, faster)
/data/scratch-oc40   (Holyoke data center, slower)

Browse archived snapshots:

ls -la /data/mydata/.zfs/snapshot

Recover a deleted file from snapshot:

cp /data/mydata/.zfs/snapshot/daily-2026-04-15/myfile.txt /data/mydata/myfile.txt

Request restore from backup:

Email help@csail.mit.edu with the full file path and approximate date range when the file last existed.

Important Reminders

Scratch storage cleanup: Files not accessed for six months are automatically deleted. Do not store permanent data, software installations, or Python environments in scratch.

Python and Conda: These will be damaged if stored in scratch because cleanup procedures delete source files while cached bytecode remains.

Use appropriate tiers: Choose the right storage tier for your use case to optimize performance, cost, and data protection.

Ask for help: Contact help@csail.mit.edu if you need assistance planning your storage or structuring your data.