Solutions
Real networks have distance, suboptimal routing, and contention working against them that are out of your control. These are the environments where Tillered makes the biggest difference, accelerating throughput without requiring changes to the underlying network, applications, or workflows.
Cloud Storage Providers
Storage providers regularly receive requests for large-scale data movement: hundreds of terabytes, cross-region or cross-continent, with fixed customer deadlines. On paper, these projects are profitable. In practice, standard WAN performance makes the timelines impossible to guarantee. So the work is declined. Not because the customer won't pay. Because the network math doesn't work.
Tillered is deployed on-demand, aligned to the project lifecycle. Deploy it for the duration of the migration. Accelerate the transfer by 10-20x. Shut it down when the work is complete.
| Metric | Baseline | With Tillered |
|---|---|---|
| Sustained throughput | 150-250 Mbps | 1-3 Gbps |
| 300 TB transfer window | 16-22 days | 6-8 days |
| Project outcome | Declined | Delivered |
| Revenue | Lost | Captured |
Why Tillered matters
When performance becomes predictable, your business changes.
You can:
- Quote fixed timelines on large migrations and meet them.
- Accept projects you previously had to decline.
- Reduce project risk without over-provisioning.
- Differentiate on delivery speed, not just price.
This is not about faster transfers. It is about turning "we can't deliver" into "yes, we can."
Enterprise IT
Enterprise teams want to run compute where power is cheap, capacity is available, or regulations allow it. The obstacle is file protocol performance. Over high-latency links, NFS and SMB fail to utilise available bandwidth, regardless of link capacity.
Render nodes sit idle. CAD tools stall. Engineers wait. The network forces compute and storage to stay co-located.
Tillered accelerates file protocol traffic over long-distance links by 20-30x. Same file servers. Same protocols. Same applications. Just fast enough for remote compute to behave like it is on the local network.
| Metric | Baseline | With Tillered |
|---|---|---|
| SMB over distance | 30-60 Mbps | 800+ Mbps |
| NFS (100 ms+) | 40-80 Mbps | 2-4 Gbps |
| Remote file access | Unusable | LAN-like |
| Compute placement | Constrained | Flexible |
Why Tillered matters
When file access performs, infrastructure decisions change.
You can:
- Place compute where power is cheapest, not where storage lives.
- Consolidate file servers without degrading remote access.
- Run render and engineering workloads across regions.
- Treat distance as a non-issue for file-based workflows.
This is not about faster file copies. It is about removing geography from compute placement.
Media & Entertainment
Modern productions want cloud workflows. What stops them is not compute, tools, or talent. It is getting footage into the cloud fast enough to keep production moving.
From remote locations, multi-terabyte daily uploads can take an entire day over standard networks. By the time footage arrives, the production day is already lost.
Tillered accelerates on-set to cloud transfers by 15-20x over long-distance links. Same upload tools. Same security. Same cloud providers. Just fast enough for the cloud to behave like a production environment instead of an archive.
| Metric | Baseline | With Tillered |
|---|---|---|
| Sustained upload speed | ~300 Mbps | 6 Gbps+ |
| 2-3 TB daily upload | ~24 hours | ~1.5 hours |
| Cloud dailies | Not viable | Same-day |
| Global collaboration | Delayed | Real-time |
Shoot in Auckland. Upload to AWS. Grade anywhere the same day.
Why Tillered matters
When uploads stop being the bottleneck, cloud economics finally work.
- Global talent becomes usable.
- Workflows scale on demand.
- Remote teams operate as if footage were local.
This is not about faster uploads. It is about making cloud post-production viable.
AI & Machine Learning
Training data is often stored in fixed regions. GPU availability and pricing vary dramatically by geography. In theory, compute should run wherever GPUs are cheapest and most available. In practice, data movement breaks the model.
Large training datasets routinely range from hundreds of gigabytes to multiple terabytes. Over long-distance links, standard TCP fails to utilise available bandwidth. Transfers that should take minutes stretch into hours.
Tillered accelerates cross-region data movement by 15-25x over high-latency links, without requiring changes to training frameworks, storage systems, or pipelines.
| Metric | Baseline | With Tillered |
|---|---|---|
| 500 GB dataset transfer | 3-4 hours | 15-20 minutes |
| 5 TB dataset transfer | 30-40 hours | 90-120 minutes |
| GPU idle time | Hours | Minutes |
| Region flexibility | Constrained | Global |
Why Tillered matters
When data moves fast enough, infrastructure decisions change.
You can:
- Train on the cheapest GPUs, regardless of where data lives.
- Iterate faster by eliminating data staging delays.
- Scale experiments without scaling transfer infrastructure.
- Treat any region as a viable compute location.
This is not incremental optimisation. It is removing geography from your AI infrastructure.
Defence & National Security
Modern operations need constant communication. Operations in outlying areas with limited comms face severe constraints. Multi-terabyte daily uploads taking an entire day over standard networks.
Tillered accelerates data transfers by 15-52x over long-distance links. Same upload tools. Same security. Same cloud providers. Demonstrated near real time delivery of high definition video enabling rapid deployment decisions.
| Route (Azure Tests) | Standard WAN | With Tillered |
|---|---|---|
| US West to AU East | 80 Mbps | 1,481 Mbps |
| Global collaboration | 91 Mbps | 4,706 Mbps |
Large file transfer between global data centres on an in-house protected network, utilising internal security management and control.
Why Tillered matters
Operational security and continuous communications are the backbone of autonomous and related Defence operations. Critical situational awareness is an ongoing need to safeguard warfighters in hostile domains.
Telemetry
Emirates Team New Zealand's race and telemetry data had to move from Catalonia to Auckland rapidly. Local ISP constraints and geographic distance limited throughput to ~5 Mbps, threatening analytics and decision-making.
Tillered re-architected the data transfer solution to maximise throughput over existing physical links while minimising workflow disruption. Same upload tools. Same security. Same cloud providers.
| Metric | Baseline | With Tillered |
|---|---|---|
| Effective throughput | ~5 Mbps | 900+ Mbps |
| Race data transfer | Multi-hour | Minutes |
| Analytics in Auckland | Delayed | Near-real-time |
Why Tillered matters
Cross-functional collaboration between design, engineering, and performance analysis teams depends on timely access to shared data. When transfer delays break that loop, decisions slow down and competitive advantage erodes.
Solutions proven in elite environments transfer directly to any organisation where distributed teams, constrained links, and time-sensitive data intersect.
Measured performance from production environments
These numbers come from production environments and controlled lab testing. In real deployments, baseline throughput is shaped by distance, BGP routing paths, retransmissions, and link contention. Tillered improved performance without changes to the underlying network.
Real-world deployments
| Route | RTT/Latency | Baseline | With Tillered | Improvement | Environment |
|---|---|---|---|---|---|
| Helsinki to Singapore NFS | ~168ms | 120 Mbps | 1.15 Gbps | 9x faster | NFS mount over AWS |
| NZ to Backblaze B2 Ashburn | ~200ms | 50 Mbps | 1.7 Gbps | 34x faster | Production Backblaze B2 |
| Apache NiFi EU to Singapore | ~160ms | 130 Mbps | 4.9 Gbps | 38x faster | Production NiFi pipeline |
| AWS S3 EU to Singapore | ~160ms | 120 Mbps | 4.7 Gbps | 39x faster | S3 standard transfer |
| Azure Sydney to US East | ~200ms | 91 Mbps | 4.7 Gbps | 52x faster | Standard Azure network |
| Geo Satellite (30 Mbps link) | ~850ms | 300 Kbps | 28 Mbps | 93x faster | High-latency satellite |
Results from production environments and controlled testing during Q3 and Q4 2025. Results vary by environment. Baseline throughput reflects real-world conditions including distance, BGP routing paths, retransmissions, and link contention. Tillered works with the existing network as-is.