Video Processing Dedicated Servers: 30% Faster Transcoding and 40% Lower Costs
About the project
European video processing platform receives source video from customers, converts it into multiple resolutions and formats, and delivers processed files to online platforms. Its workload combines sustained CPU utilization with large volumes of incoming and outgoing traffic.
Challenge
The platform originally used a combination of VPS instances and public cloud services. As processing volumes increased, two problems became increasingly important. First, CPU performance during long transcoding jobs was not always consistent. Second, bandwidth became a significant infrastructure expense due to the volume of large video files moving between systems.
The platform required:
- high single-core and multi-core CPU performance;
- NVMe storage;
- 10 Gbps connectivity;
- high traffic allowances;
- predictable bandwidth costs;
- simple horizontal scaling.
Solution
The processing infrastructure was gradually migrated to a pool of Unihost dedicated servers.
KDS-04 – AMD Ryzen 9 9950X: 16 cores, 64 GB DDR5 RAM, 960 GB NVMe drives, 10 Gbps, 330 TB traffic.
Used for video transcoding, media processing, temporary file processing, encoding queues, and high-bandwidth data transfer.
Multiple KDS-04 servers operate as independent processing nodes. A job scheduler distributes encoding tasks between available servers according to current utilization. A separate node handles temporary storage and processing queues. The architecture can therefore be expanded simply by adding another processing server to the pool.
Results
After moving the primary processing workload to dedicated infrastructure, average processing time for typical video jobs decreased by approximately 25 – 30%. Total compute and bandwidth expenses decreased by approximately 40% compared with the previous infrastructure at a similar workload level. The business also gained more predictable project economics. Instead of treating bandwidth as a highly variable expense, the team can estimate the infrastructure requirements of a large video-processing project before accepting it.
What’s next
The platform plans to expand its processing cluster and deploy additional storage and backup capacity. Results at a glance: pp to 30% faster video processing, approximately 40% lower infrastructure costs, 10 Gbps connectivity, scalable processing cluster.
