This content was auto-generated from Fusion SMB documentation and is pending SME review. Please verify accuracy before using in partner-facing contexts.
Fusion SMB — Value Proposition
This page outlines the key reasons customers choose Fusion SMB, organized around the value pillars that resonate most in sales conversations.
1. Unmatched Performance
Fusion SMB is the fastest SMB server available on Linux, purpose-built for throughput-intensive workloads.
| Metric | Fusion SMB | Samba (Open Source) |
|---|---|---|
| Max throughput | ~25 GB/s (customers have saturated 400 Gbps links) | ~2.5 GB/s |
| Architecture | Multi-threaded, tunable | Process-based (1990s, single-threaded per connection) |
| Concurrent active workload connections | 10,000s | 100s |
| Small file creates / sec | 1,000s | 10s |
| Small file writes / sec | 10,000s | 100s |
| SMB Direct (RDMA) | ✅ Supported | ❌ Not supported |
| SMB Multichannel | ✅ Supported | Limited support |
| SMB compression | ✅ Supported | ❌ Not supported |
| Zero-copy reads | ✅ Supported | ❌ Not supported |
Single-Client Benchmark Numbers
Concrete throughput on a single client with FIO 8MB IO, no buffering, latest software:
| Network | Fusion SMB | Samba |
|---|---|---|
| 100 GbE (RDMA) | 11.4 GB/s | 2.4 GB/s |
| 200 GbE | 20.4 GB/s | 2.44 GB/s |
Server-Side Resource Profile
Running 2,500 simultaneous active SMB clients on an AWS c5.2xlarge VM (8 vCPU Xeon 3.00GHz, 16GB mem) over TCP (RDMA would be even lower CPU):
| Test | CPU max | CPU avg | Mem max | Mem avg |
|---|---|---|---|---|
| 1M IO block, read | 19.14% | 8.72% | 3.5 GB | 2.4 GB |
| 1M IO block, write | 30.20% | 13.43% | 3.2 GB | 2.1 GB |
| 4K IO block, read | 58.41% | 49.99% | 175 MB | 174 MB |
| 4K IO block, write | 59.04% | 49.36% | 228 MB | 219 MB |
| 8K IO block, read | 57.81% | 51.22% | 177 MB | 175 MB |
| 8K IO block, write | 61.55% | 50.24% | 229 MB | 221 MB |
Reliability Under Load — 100% vs Samba's 49% Failure Rate
When Samba is pushed past its process-per-connection ceiling, it doesn't slow down — it drops clients. In a sustained random 4K write test:
| Server | Clients sustained |
|---|---|
| Fusion SMB | 100% success |
| Samba | 51% success / 49% failure |
Fusion SMB keeps working when Samba has created too many processes for its cores and starts failing clients outright.
Test setup: Lenovo P52s mobile workstation (8th gen i7-8650U, 32GB DDR4, NVMe), Ubuntu Linux, Samba 4.8.0.
Up to 61× Small-File Creation Advantage
Samba's per-connection process model is particularly punishing for small-file workloads. Using Oracle vdbench to create 30,000 1KB files:
| Workload | Fusion SMB | Samba | Advantage |
|---|---|---|---|
| Small file creation (standalone) | 2,307 files/sec | 40 files/sec | 57× |
| Small file creation (clustered) | 2,000 files/sec | 33 files/sec | 61× |
| Small file random write (standalone) | 8,842 files/sec | 983 files/sec | 9× |
| Small file random write (clustered) | 6,231 files/sec | 448 files/sec | 14× |
This matters for build systems, version control servers, application package distribution, and any directory-heavy workload.
Container Performance — Fusion in a Container Beats Samba on Bare Metal
Running TCP single-thread robocopy (1,400 files, 21 GB total, ~15 MB avg):
| Configuration | Throughput |
|---|---|
| Fusion (host) | ~612 MB/s |
| Fusion (container) | ~608 MB/s |
| Samba (host) | ~588 MB/s |
| Windows Server 2025 | ~562 MB/s |
| Samba (container) | ~553 MB/s |
A containerized Fusion deployment outperforms Samba running directly on the host — and Windows Server, and a containerized Samba. Containerization isn't a tradeoff for Fusion; it's a deployment option with zero performance penalty.
Sustainment, small-file, and container performance numbers are sourced from the SMB + NFS engineering deck. Confirm publication-readiness before quoting in partner-facing materials.
SMB Compression — Dramatic Wins on Bandwidth-Constrained Links
Copying a 4 GB VHDX with 250 MB of uncompressible JPGs over a typical cable broadband link:
| Configuration | Time to copy |
|---|---|
| Samba | 1,645 seconds |
| Samba (with compression requested) | 1,669 seconds (Samba has no compression — request is ignored) |
| Fusion SMB (no compression) | 1,563 seconds |
| Fusion SMB (compression requested) | 360 seconds — 4.6× faster |
Single-client, 2,500-client, and compression benchmark numbers are sourced from the Fusion SMB Battlecard. Confirm publication-readiness before quoting in partner-facing materials.
ML/AI Workload Performance
Fusion SMB demonstrates significant advantages on ML/AI workloads. Recent SNIA Developer Conference 2025 NFS vs SMB head-to-head comparison (RDMA, dual port):
| Test | NFS | SMB (Fusion) |
|---|---|---|
| MLPerf 3D-unet | 14 @ 93% | 15 @ 93% |
| MLPerf checkpointing | 16.2 / 18.5 GBps | 23.5 / 28.2 GBps |
| FIO sequential WRITE / READ | 29.2 / 42.2 GBps | 48.4 / 48.8 GBps |
- MLPerf 3D-unet training: Fusion SMB outperforms NFSD, Ganesha, and Samba (Ganesha and Samba limited to TCP — no RDMA support)
- MLPerf llama3-8b checkpoint: Fusion SMB delivers faster checkpoint writes critical for large language model training
- FIO comparison: Approaches maximum storage write speed (22 GB/s on 8×RAID-0 NVMe XFS)
MLPerf benchmark results are from Aaron Kennedy's IBM UG presentation (April 2026) and the Fusion SMB Battlecard. Confirm specific numbers and whether they can be cited in partner-facing materials.
Why It Matters to Customers
- Media & entertainment — 4K/8K video editing workflows require sustained high-throughput access to shared storage
- Engineering & CAD — large assembly files and simulation data sets need low-latency file access
- Healthcare & life sciences — medical imaging (DICOM) and genomics data demand high bandwidth
- Financial services — high-frequency data ingestion and analytics on shared file stores
Performance Tunability
Unlike one-size-fits-all solutions, Fusion SMB provides granular performance tuning:
- Configurable network and I/O thread pools (
transport_rx_threads,transport_tx_threads,vfs_data_threads) - Per-share synchronous I/O and buffering controls
- File space preallocation for write-heavy workloads
- Configurable SMB2 credit limits and message sizes
- Connection and session limits for capacity planning
2. Business Continuity & High Availability
Fusion SMB provides three levels of resilience, allowing customers to match their HA strategy to their requirements:
Continuous Availability (CA)
- File shares remain accessible during planned maintenance (server updates, rolling upgrades)
- Persistent file handles enable transparent client reconnection
Active-Passive Failover
- Standby node takes over automatically on primary failure
- Floating IP address ensures seamless client redirection
- Works with standard file systems (ext4, XFS, ZFS, NTFS)
Active-Active Scale-Out
- Multiple active nodes share the client load
- Automatic state replication across nodes
- Supports clustered file systems (GlusterFS, WekaFS, CephFS, Lustre, GPFS, GFS2, OCFS2)
- Clients transparently rebalance after node failure or addition
Rolling Upgrade — Zero Downtime
Fusion SMB supports rolling cluster upgrades with zero downtime:
- Stop Fusion service on one node — clients automatically move to remaining nodes
- Update the Fusion package on that node
- Repeat for each node in the cluster
- Upon the last node upgrade, all nodes automatically enter the new functionality mode
- No node eviction required
- Supports n-2 version compatibility during the rolling process
- Clients get new features dynamically or on next connection
- Simplest change control, lowest operational overhead
Infrastructure Integration
- Uses industry-standard clustering software (Corosync + Pacemaker)
- TCP Tickle ACK mechanism for fast client recovery
- Network separation for cluster and client traffic
- Compatible with existing enterprise HA infrastructure
3. Microsoft Partnership & Protocol Compliance
Tuxera is an official Microsoft SMB technology partner and patent licensee. This means:
- Full protocol compliance — not a reverse-engineered implementation
- Access to Microsoft SMB specifications and test suites
- Ongoing compatibility as Microsoft evolves the SMB protocol
- Reduced integration risk for customers connecting Windows clients to Linux storage
4. Total Cost of Ownership (TCO)
Linux Infrastructure Savings
- Eliminates Windows Server licensing costs for file serving
- Runs on commodity Linux hardware
- Small footprint — fewer resources consumed than Windows File Server
Operational Efficiency
- Centralized configuration via a single config file
- Package-based deployment (deb/rpm) for standard Linux workflows
- CLI tools for monitoring and diagnostics
- Can coexist with NFS on the same server — no dedicated hardware needed
Scalability Without Forklift Upgrades
- Add nodes to an active-active cluster for linear capacity growth
- No vendor lock-in to proprietary storage hardware
- Works with customer-chosen file systems and storage backends
5. Security & Compliance
- Enterprise authentication via Kerberos and Active Directory
- SMB encryption (AES-128/256, AES-256-GCM) for data in transit
- Windows ACLs for permission management familiar to IT admins
- Access-Based Enumeration (ABE) for privacy and access control
- Audit logging for regulatory compliance (SOX, HIPAA, GDPR data access tracking)
Industry-Specific Value
Different verticals respond to different framings. Use these tailored pitches when leading with relevance:
Cloud Solution Providers
- Compatibility & safety — enterprise-level speed across Windows, Linux, and macOS, ideal for heterogeneous workloads and VDI (Azure Virtual Desktop, Amazon WorkSpaces, Google Virtual Desktop). Direct IO and persistent handles guarantee data integrity.
- Lowest overhead, greatest scale — multi-threaded user-mode service doesn't spawn multiple processes (unlike Samba), freeing memory and CPU for application workloads. Scales effortlessly from tiny containers to 32-node clusters.
- Enterprise-grade features — SMB Direct, compression, and large clusters Samba can't match. Tens of thousands of clients pushing massive data through your network. Ideal as part of a PaaS or SaaS offering.
Data Platform Providers
- Maximum throughput for massive datasets — removes the Samba bottleneck slowing customer storage. Full RDMA and SMB Direct mean customers can buy your most high-end equipment and see it fully utilized.
- Scalable data integrity — deploy in containers and Kubernetes as part of microservice fleets. Sell 32-node scale-out clusters with Direct IO and persistent handles guaranteeing integrity through temporary network failures.
- Direct support from the experts — talk directly to escalation engineers and developers, not forums or outsourced tier 1. Tuxera's product manager spent 12 years at Microsoft architecting SMB 3.
System Integrators
- Eliminate Samba bottlenecks — full performance potential delivered via RDMA and SMB Direct. Customers can buy your high-end storage and network recommendations and see them fully utilized.
- Cross-platform compatibility — modern SMB 3 features for Windows, macOS, and Linux. Unlike NFS 4 (which leaves Windows out), every client gets the same secure, high-performance experience.
- Direct support from the experts — global 24x7 with engineers backed by code developers.
Media & Entertainment
- Unmatched throughput — up to 12 GB/s on Windows and 4.1 GB/s on macOS per 100 Gb network. Real-time editing of 4K and 8K content where multiple workstations need maximum shared performance. Avoid buying extra equipment to compensate for Samba's limitations.
- Flexible deployment — on-prem, clusters, containers, or cloud. Support remote teams, integrate with existing workflows, scale from single servers to 32-node scale-out clusters as production evolves.
- Enterprise security & support — content protection with encryption, ABE, and Kerberos. Connects to Active Directory, LDAP, and Apple Open Directory for secure deployment.
Medical & Healthcare
- Secure data transfers — AES-256-GCM encryption for cybersecurity threat prevention and legally-mandated confidentiality compliance across connected healthcare systems. ABE protects patient privacy.
- Reliable technology — Direct IO and persistent handles (both absent from Samba) guarantee data and system integrity during unexpected power outages, protecting critical patient information.
- Streamline access to large medical datasets — super-resolution microscopes, digital pathology, and imaging devices generate huge files. Transfer and access at up to 12 GB/s on Windows and 8.6 GB/s on macOS per 100 Gb network.
The following areas would strengthen this page with real data:
- Customer benchmark results — anonymized performance comparisons from actual deployments
- TCO calculator or model comparing Fusion SMB vs. Windows File Server licensing over 3-5 years
- ROI examples from existing customers
For performance tuning details, see Fusion SMB Documentation — Performance Tuning. For clustering details, see Fusion SMB Documentation — Clustering.