AI Summary
Premier Care includes 500 GB of backup storage on Dedicated Servers and 2 TB on High Capacity Servers, which covers an off-server copy of important recordings.
Every managed dedicated server includes Advanced Product Support, and all tiers above entry level can add Premier Care for malware defense, backup storage, and consulting hours.
What Does a Streaming Server Cost Per Month?.
Basic summary
Most streaming buyers overspend on CPU and underspend on network capacity. This guide walks through how to size a dedicated server for streaming based on the numbers that actually govern video quality: cores per rendition, port speed against concurrent viewers, storage growth from recordings, and the point where a CDN should take over delivery.
What Should You Size First on a Streaming Server?
Port speed. Everything else is negotiable.
A streaming server runs three jobs with completely different resource profiles. Ingest accepts the encoded feed from your broadcast software and costs almost nothing. Transcoding rebuilds that feed into multiple quality levels and consumes CPU. Delivery pushes segments to viewers and consumes network capacity, which is where budgets fail first.
You can compensate for a slow CPU by encoding fewer renditions. You can compensate for small disks by archiving to object storage. You cannot compensate for a saturated network port. Once the port is full, every viewer buffers at the same time, and the only fix is a hardware or plan change.
That ordering shapes the rest of this guide: network first, CPU second, RAM third, storage last.
How Do You Calculate Bandwidth for Concurrent Viewers?
Multiply concurrent viewers by their average delivered bitrate, then add overhead. A 1080p stream at 5 Mbps and a 720p stream at 3 Mbps are reasonable planning figures for adaptive delivery where most viewers land in the middle of the ladder.
Two adjustments keep this honest. Port speed is the rated capacity of the connection, not the throughput you will measure. Protocol overhead, retransmits, and network conditions all take a cut, so plan against roughly 60 to 70% of the rated figure for sustained delivery. Second, concurrency is not audience size. A stream with 4,000 registrations and a 35% live attendance rate peaks near 1,400 viewers, not 4,000.
InMotion’s High Capacity Servers ship with a 3Gbps base port and unmetered transfer, with additional capacity available in 1Gbps increments at $60 per month up to 10Gbps. The CC-4000 includes the full 10 Gbps at no premium. Unmetered matters as much as the port rating here, since metered plans turn a successful stream into an overage invoice.
How Much CPU Do You Need for Live Transcoding?
Budget by rendition, not by stream.
Software encoding with x264 at the veryfast preset consumes roughly 2 to 4 cores per 1080p30 rendition, depending on scene complexity. A three-rung ladder (1080p, 720p, 480p) generated from a single 1080p source generally lands between 6 and 10 cores once ingest and packaging overhead are counted. Two simultaneous channels roughly double that.
Preset choice moves the number more than hardware does. FFmpeg’s H.264 encoding guide explains the tradeoff: slower presets buy compression efficiency at a steep CPU cost, which is fine for on-demand encoding and unworkable for live. Live transcoding has a hard deadline. If encoding falls behind real time, frames drop, and no amount of CPU added later repairs the recording.
Three ways to reduce the CPU bill before buying more of it:
Pass through the top rung. Deliver the source 1080p feed untouched and transcode only the lower rungs. This alone removes the most expensive rendition from the workload.
Trim the ladder. Four rungs are common. Three are usually enough for a business audience on broadband and desktop devices.
Encode 30 fps instead of 60 fps for talking-head content. Sports, gameplay, and motion-heavy footage need the higher frame rate. Webinars and product demos do not.
InMotion does not offer GPU servers, so hardware-accelerated encoding is not part of this stack. Software encoding on modern cores handles single-channel and small multi-channel work comfortably; very large ladders across many simultaneous channels belong on a dedicated encoding tier or an external transcoding service.
How Much RAM Does a Streaming Server Actually Use?
Less than most buyers expect for a single channel, and much more once you run several.
The streaming stack itself is modest. NGINX with the RTMP module, an FFmpeg process per rendition, and the operating system rarely exceed 8 to 16 GB for one channel. That surprises a lot of first-time buyers who assume video is memory-hungry the way databases are.
RAM starts to matter when the server acts as an origin. Segments served from page cache never touch disk, which keeps latency flat under load. A single 5 Mbps rendition with a 60 second DVR window holds about 37 MB in cache. Multiply that across a four-rung ladder and twenty channels and you are caching several gigabytes of live segments continuously, on top of the encoding processes.
The 64 GB configurations on InMotion’s Managed Dedicated Servers leave enough headroom for a multi-channel origin without forcing segment reads back onto disk.
What Storage Configuration Fits Video Delivery Infrastructure?
NVMe for active segments, cheaper capacity or object storage for the archive.
Live segments are small files written and deleted constantly. That write pattern rewards NVMe and punishes spinning disks. Recordings are a different problem: they only grow.
Recording math is predictable. A 1080p capture at 6 Mbps consumes about 2.7 GB per hour. At 12 Mbps, closer to 5.4 GB per hour.
Eight hours per day at 6 Mbps: roughly 22 GB daily, or 650 GB per month
Dual 1.92 TB NVMe in RAID-1 gives usable capacity near 1.92 TB, so that schedule fills the array in about three months
RAID-1 mirrors for redundancy, so a two-drive array does not double usable space
Plan the archival path before the first stream, not after the disk alert. Premier Care includes 500 GB of backup storage on Dedicated Servers and 2 TB on High Capacity Servers, which covers an off-server copy of important recordings. For long-term libraries, sync completed recordings to S3-compatible object storage on a schedule and keep the local disk for the working set.
Which Protocol Should You Deliver In?
Latency and scale pull in opposite directions, and the right answer depends on which one your audience actually needs.
Apple’s HTTP Live Streaming documentation covers the Low-Latency HLS extension, which uses partial segments to cut delay while keeping the HTTP delivery model that makes CDN caching possible. That caching property is the whole reason HLS scales cheaply: a segment is a static file, so one server can feed a CDN that feeds everyone else.
WebRTC holds a stateful connection per viewer. Server cost per viewer rises sharply as a result, which is why interactive platforms typically run WebRTC for the handful of participants who speak and HLS for the audience that watches.
When Should a CDN Handle Delivery Instead of Your Server?
Once sustained delivery approaches about half your usable port capacity, or once your audience spans more than one continent.
The economics are clean. Without a CDN, 2,000 concurrent viewers at 5 Mbps pull 10 Gbps directly from your server. With a CDN in front, the origin serves each segment once to each edge location and sees a few dozen connections instead of 2,000. Your server load stops scaling with audience size.
Geography forces the same decision at lower viewer counts. A viewer in Sydney pulling segments from a Virginia origin waits for every round trip. InMotion operates data centers in Virginia, California, and Amsterdam, which shortens the first hop considerably for North American and European audiences, but a global audience still needs edge caching.
Configure the origin correctly before pointing a CDN at it. Segment files should carry long cache lifetimes since they never change once written. Playlist manifests should carry very short ones, because they change every few seconds. Getting that backwards produces either stale streams or a CDN that caches nothing.
Does On-Demand Video Change the Requirements?
Substantially. VOD removes the real-time deadline that drives streaming server hardware decisions.
Encoding a library overnight can use slow, efficient presets that produce smaller files at the same visual quality. Smaller files mean lower delivery bandwidth for every future view, so the CPU spent once pays back continuously. Live encoding never gets that option.
What VOD adds instead is storage. A 500 hour library at 1080p across a three-rung ladder runs into several terabytes, and libraries only grow. If your workload is mostly on-demand, weight the budget toward capacity and CDN offload rather than cores.
Which InMotion Server Fits Your Streaming Workload?
Every managed dedicated server includes Advanced Product Support, and all tiers above entry level can add Premier Care for malware defense, backup storage, and consulting hours.
What Does a Streaming Server Cost Per Month?
The server line item is rarely the whole bill. Budget for four components:
The server itself, from $99.99 per month for a configuration that handles a single well-tuned channel
Port speed above the base, at $60 per month per additional gigabit on plans where that add-on applies
CDN delivery, billed per terabyte by your provider and often the largest variable cost at scale
Archival storage, either included backup capacity or object storage priced per terabyte
A control panel license, where you need one, is billed separately from managed care bundles. Confirm current pricing in the cart before committing, since promotional and renewal rates differ.
What Causes Buffering After You Have Bought the Server?
Hardware is rarely the culprit. These five configuration problems account for most post-purchase complaints:
Keyframe intervals that do not align with segment length. Players cannot switch renditions cleanly, so viewers see stalls during quality changes. Set the keyframe interval to twice the frame rate and match it across every rung.
A preset that is too slow for live. Encoding falls behind real time and frames drop. Reach for veryfast and adjust from there.
RTMP left open to the internet. Anyone who finds the port can push a stream through your bandwidth. Restrict ingest to known broadcaster addresses.
Recordings written to the same volume as live segments. Disk contention shows up as stutter during long sessions. Separate the paths.
No CDN in front of an origin serving a global audience. Distance is not a hardware problem, and no server upgrade fixes it.
Where Should You Get a Second Opinion on the Build?
Streaming workloads vary more than almost any other hosting use case. A single 720p webinar channel and a four-channel 1080p60 sports origin sit at opposite ends of the same product line, and the difference between a right-sized build and an expensive mistake usually comes down to two numbers: peak concurrent viewers and how many renditions you actually need.
If those numbers are still estimates, talk them through before ordering. InMotion’s team works with streaming and media workloads across shared origin setups, multi-channel builds, and CDN-fronted architectures, and sizing conversations happen with engineers rather than a script. Contact the team with your concurrency estimate and encoding ladder, or review current Dedicated Server plans to compare configurations directly.