Advantages describe what MooseFS does for your business — why your CFO, CTO and operations team should care. Features describe how it does it — the technical capabilities your engineers will scope, deploy and operate.
Designed for mission-critical applications. No single point of failure, redundant metadata, and automatic failover. The cluster keeps serving while broken disks are detected and re-replicated in the background.
Every disk is monitored for health. Failing media is marked damaged and affected chunks are replicated to healthy nodes automatically — no operator action required. Recovery speed scales with the cluster: more chunkservers mean more replication sources and destinations, so a larger or faster cluster recovers a failed drive proportionally quicker.
Atomic snapshots, a configurable per-directory trash bin, and an off-master metalogger backup turn the usual “I deleted the wrong thing” or “the master crashed” emergencies into routine recoveries rather than tape restores.
Files are striped across chunkservers and served in parallel; clients talk directly to storage nodes. Aggregate bandwidth tracks the number of chunkservers — add nodes, get more throughput.
MooseFS mounts as an ordinary filesystem with full POSIX semantics on Linux, FreeBSD and macOS. Applications need zero changes — and your operators use the Unix permissions and ACLs they already understand.
Schedule jobs directly on the storage nodes you have already deployed. Idle CPU and memory on chunkservers become a compute resource — no separate fleet to buy, run, or pay licenses on.
Open-source code, commodity x86-64 or ARM hardware, and erasure coding that reclaims more than half the raw space a three-copy setup would burn. Predictable capex, predictable TCO, no vendor lock-in.
MooseFS Pro is licensed once for the capacity you buy — not as a recurring annual subscription. Pay one time for the TBs you need and keep running them for as long as your hardware lasts. Renewals are for support, not for the right to keep using the software.
Start free with the Community Edition for evaluation, testing or non-critical workloads. Move to Pro when you need high availability, enterprise features and support.
License and support pricing improves as your cluster grows. The customers who use MooseFS most pay the least per terabyte of protected data — no penalty for success.
Mix NVMe, SSD and HDD across tiers. Put older, smaller and slower servers in less intensive tiers — they still earn their keep. New hardware joins the cluster without a forklift upgrade, on x86-64 or ARM from any vendor.
A single instance addresses up to 16 EiB and 2.1 billion files. Add a chunkserver and capacity, redundancy and performance scale with it — no downtime, no migration, no metadata panic.
Hardware replacements, software upgrades and capacity additions, one node at a time. Your platform stays current while your cluster stays online — no maintenance windows, no migration projects.
VM disks, AI training data, video archives, surveillance recordings — all in a single cluster with per-directory quotas, ACLs and redundancy policies. One pool to monitor, one bill, one tool to learn.
The Community Edition is released under GPLv2 — open code, no surprises. Audit it, modify it, run it on as many machines as you need. Pro adds enterprise features on top, but never holds the basics hostage.
Running real workloads worldwide for twenty years — internet TV broadcasters, research clusters, CDN origins, financial archives. The kind of platform that has had time to have the bugs fixed.
Global support direct from the MooseFS engineering team — by email and phone, the same people who write the code.
All system components are redundant — metadata, chunkservers, network paths. An automatic failover mechanism keeps the namespace online during any single component failure, and the client never sees it.
Instantaneous, uninterrupted point-in-time copies of any directory tree. Ideal for online backup solutions and safe migrations — no application pause, no file copy.
MooseFS checks every disk for health, marks failing media as damaged, and replicates affected chunks to healthy nodes automatically. A 2 TiB drive failure replicates from another copy in about 40–60 minutes.
Set the number of copies (the goal) per file or directory. Critical data can be triplicated, bulk archive can run with a lighter goal — all inside one shared namespace.
A virtual, global space for deleted objects, configurable for each file and directory. Recover accidentally deleted data without a tape pull, with a per-path retention window.
A separate daemon pulls metadata from the master at intervals and stores it for backup. The namespace can be reconstructed even if the master server’s disks are lost.
Reads and writes are served in parallel across every chunkserver involved. Clients talk directly to chunkservers — the master is never on the data path.
Large files are divided into chunks up to 64 MB, distributed across nodes and served concurrently — aggregate bandwidth is the sum of every node involved.
Exceptional performance on almost every hardware platform that runs a POSIX-compliant operating system. Applications use it without changes — it mounts as an ordinary filesystem.
Dedicated FUSE-based client (mount) components specially designed for Linux, FreeBSD and macOS systems deliver high performance straight from the kernel.
A separate native MooseFS Client for Microsoft Windows built on top of Dokany. Mount the cluster as a drive on Windows servers and workstations.
Schedule computation directly on data nodes to improve overall TCO by using idle CPU and memory. No need to buy extra hardware — the same server stores data and runs applications.
4+n and 8+n schemes with up to nine parity sums. Typically saves 55%+ of raw space versus three-copy replication while keeping the same fault tolerance.
Assign categories of data to media types. Hot data on fast SSD, infrequently used data on cheaper, slower mechanical drives — all governed by storage classes.
Build on x86-64 or ARM from nearly any manufacturer. Mix older, smaller and slower servers in less intensive tiers — no vendor lock-in, no uniform-drive requirement.
A free Community Edition and a one-time Pro license that scales with your cluster. Support and license discounts deepen as the storage size grows.
Up to 16 EiB of capacity and over 2.1 billion files in a single instance, with individual files up to 128 PiB. One namespace, no sharding required.
Perform one-node-at-a-time upgrades, hardware replacements and additions without disruption of service. Keep the platform up to date with no downtime.
Access based on a standard Unix access-control model enhanced with standard Access Control Lists — the way your operators already work.
Per-directory storage capacity limits. Keep tenants, projects and teams predictable inside one shared namespace.
A rich set of administrative tools — command-line and web-based interfaces — for monitoring, diagnostics, capacity analysis and configuration. Built for operators.
