Edge 10G
10 GbpsAmsterdam, NL
- CPU
- 16c / 32t
- Model
- AMD EPYC 7313P
- Memory
- 128 GB ECC
- Storage
- 2 × 1.92 TB NVMe
- Traffic
- Unmetered
- IPv4
- 4 included
€329.00/ month
No setup fee
AS208220 · Streaming infrastructure
IPTV and OTT servers, built around the network
Video is a bandwidth business. These builds pair unmetered 10–25 Gbps ports on our own autonomous system with the CPU and NVMe a live ABR ladder actually needs.
AS208220Unmetered ports10 – 25 GbpsDDoS filteredEurope · Asia
Autonomous system
AS208220
Offerhost
Network availability
99.990%
Rolling 30 days
Delivery locations
5
Europe and Asia
Port capacity per build
10 – 25 Gbps
Unmetered
Demo data. Network availability shown here comes from the mock provider. No live monitoring backend is connected, so these values are generated placeholders.
Everything on this page rides AS208220 — the same network our status page, looking glass and peering policy describe. Run a traceroute from any of our locations to your audience before you order. We would rather you check than take our word for it.
Architecture
From encoder to set-top box
Run every stage on us, or drop us in as the edge in front of an origin you already operate.
Delivery chain
Ingest → screen
Ingest
Your source
SRT, RTMP or MPEG-TS arrives from your encoder or satellite receiver.
SRTRTMPMPEG-TSTranscode
Bitrate ladder
One mezzanine input becomes an ABR ladder so every device gets a stream it can play.
H.264HEVCABR ladderOrigin
Packaging and DVR
Segments packaged to HLS or DASH, with catch-up and VOD held on NVMe.
HLSDASHCatch-upEdge
Delivery
Unmetered ports on AS208220, close to your audience and peered locally.
Unmetered10–25 GbpsAnycast DNSViewers
Playback
Set-top boxes, smart TVs, mobile apps and browsers.
STBSmart TVMobile
Sizing
Work out what you actually need
Concurrent viewers multiplied by bitrate is the number that picks your server. Here is the arithmetic, with the reasoning shown.
Capacity planner
10050,000
The usual default
- 2,000 × 6 Mbps
- 12.00 Gbps
- With peak headroom
- 15.60 Gbps
Suggested build
2 × Edge 10G
- Total port capacity
- 20 Gbps
- Per server
- €329.00
- CPU
- AMD EPYC 7313P
- Memory
- 128 GB
- Traffic
- Unmetered
€658.00/ month
≈ €0.33 per concurrent viewer
A planning estimate, not a quote. Real capacity depends on your bitrate ladder, ABR switching behaviour, cache hit ratio and how concentrated your peak is. Send us your numbers and we will size it properly.
Builds
Streaming configurations
Four starting points. Any of them can be adjusted — these are the specs we keep parts for.
Order Edge 10G
Origin 10G
10 GbpsFrankfurt, DE
- CPU
- 24c / 48t
- Model
- AMD EPYC 7443P
- Memory
- 256 GB ECC
- Storage
- 4 × 3.84 TB NVMe
- Traffic
- Unmetered
- IPv4
- 8 included
€489.00/ month
No setup fee
Transcode 25G
25 GbpsAmsterdam, NL
- CPU
- 32c / 64t
- Model
- AMD EPYC 9354P
- Memory
- 384 GB ECC
- Storage
- 4 × 3.84 TB NVMe
- Traffic
- Unmetered
- IPv4
- 8 included
€749.00/ month
No setup fee
Edge 10G APAC
10 GbpsSingapore, SG
- CPU
- 32c / 64t
- Model
- Intel Xeon Gold 6338
- Memory
- 256 GB ECC
- Storage
- 2 × 3.84 TB NVMe
- Traffic
- Unmetered
- IPv4
- 4 included
€579.00/ month
No setup fee
| Build | Processor | Cores | RAM | Storage | Port | Traffic | Location | Price |
|---|---|---|---|---|---|---|---|---|
| Edge 10G | AMD EPYC 7313P | 16c / 32t | 128 GB | 2 × 1.92 TB NVMe | 10 Gbps | Unmetered | Amsterdam, NL | €329.00 |
| Origin 10G | AMD EPYC 7443P | 24c / 48t | 256 GB | 4 × 3.84 TB NVMe | 10 Gbps | Unmetered | Frankfurt, DE | €489.00 |
| Transcode 25G | AMD EPYC 9354P | 32c / 64t | 384 GB | 4 × 3.84 TB NVMe | 25 Gbps | Unmetered | Amsterdam, NL | €749.00 |
| Edge 10G APAC | Intel Xeon Gold 6338 | 32c / 64t | 256 GB | 2 × 3.84 TB NVMe | 10 Gbps | Unmetered | Singapore, SG | €579.00 |
55 Gbps of port capacity across the four standard builds. Need a different shape? Build it in the configurator or ask for a quote.
Why the network matters
Buffering is a routing problem more often than a CPU problem
Our own AS number
Streaming lives or dies on the network. AS208220 is ours, so routing decisions, congestion and peering are things we fix rather than escalate.
Unmetered ports
Priced per port, not per terabyte. A viewership spike costs you nothing extra and never triggers an overage invoice mid-month.
Filtering that keeps you on air
Volumetric attacks are scrubbed rather than null-routed. Streaming platforms are attacked constantly and a null route is an outage by another name.
Transcoding-grade CPUs
High-core EPYC and Xeon parts sized for live ABR ladders, with NVMe for segment cache, catch-up and VOD.
Placement
Put the edge where the audience is
Latency follows geography. Serving Indian subscribers from Amsterdam works, but Singapore works better.
Amsterdamams
Netherlands
Western European entry point with short paths to the major exchanges.
Frankfurtfra
Germany
Central European hub for customers serving DACH and CEE traffic.
Londonlon
United Kingdom
UK and Ireland facing capacity for latency-sensitive workloads.
Warsawwaw
Poland
Planned build-out. Register interest to be notified when it opens.
Singaporesin
Singapore
South-East Asian presence for APAC-facing applications.
Tokyotyo
Japan
North-East Asian capacity with direct routes toward the Pacific.
Hong Konghkg
Hong Kong SAR
Planned build-out serving Greater China routes.
Getting started
How a launch usually goes
- 1
Send us your numbers
Channel count, bitrate ladder, expected concurrency and where your audience is. That is enough for us to size it.
- 2
Test the route first
Run a traceroute from the location we suggest to a few of your subscribers. If the path is bad, we would both rather know now.
- 3
One origin, then edges
Most platforms start with a single origin and add regional edge servers once the audience concentrates somewhere.
- 4
Scale per event
Tell us ahead of a known spike. Adding a server mid-month is prorated, and we watch capacity with you on the night.
Delivery times
- In-stock build
- A few hours
- Custom spec
- 2 – 5 working days
- Extra edge server
- Same day where in stock
What you run
Unmanaged bare metal with full root and IPMI. Bring nginx-rtmp, SRS, Wowza, Flussonic, your own packager — whatever your stack is. We do not install or support the streaming software, and we do not charge you for a panel you did not ask for.
Before you commit
Ask for a test IP in the location you are considering and run your own load test against it. Any host confident in their network will say yes to that.
Who these servers are for
Broadcasters, OTT platforms, sports and event streamers, houses of worship, education providers, and IPTV operators distributing content they hold the rights to. If you are licensed to distribute it, we are glad to carry it, and we will size the network around your peak rather than your average.
What we will not host is redistribution of channels you have no rights to. That is covered in our acceptable use policy, and we enforce it — partly because it is the law, and partly because rights complaints get networks depeered, which would degrade service for every legitimate streamer on AS208220. We would rather be straight about that before you order than after.
Questions
The things people ask first
Unmetered really means unmetered?+
Yes — you pay for the port speed, not the traffic that crosses it. A 10 Gbps unmetered port can move roughly 3.2 PB in a month if you saturate it. What we do ask is that you tell us before a known spike, such as a big match or a launch, so we can watch capacity with you.
How many viewers fits on one server?+
Bandwidth is almost always the limit, not CPU. At 6 Mbps for 1080p, a 10 Gbps port carries roughly 1,600 concurrent viewers with peak headroom, and a 25 Gbps port around 4,000. The planner above does this arithmetic for your own bitrate ladder.
Can you transcode as well as deliver?+
Yes. Software transcoding on the high-core builds handles live ABR ladders; if you need GPU encoding for a dense channel count, tell us the channel and profile count and we will spec the card with you.
Do you provide the panel or the software?+
No — the servers are unmanaged bare metal, so you run your own stack. Most customers deploy nginx-rtmp, SRS, Wowza, Flussonic or their own packager. We handle the metal and the network.
How fast is delivery?+
In-stock configurations are typically racked within a few hours. Custom builds with unusual storage or GPUs take two to five working days. We confirm by email before anything is charged.
Can I add capacity mid-month?+
Yes, and it is prorated. Most platforms start with a single origin, then add edge servers per region as the audience concentrates somewhere specific.
AS208220
Tell us your concurrency, we’ll size the network.
Channel count, bitrate ladder and where your viewers are. You will get a spec and a price, not a discovery call.