Skip to main content

System Requirements - Load Baance Server

Below is a rough guideline for server hardware requirements for a transcodingload balance server

Assumptions:

  • Stream Bitrate: We’ll assume each “complete” stream (all ABR renditions combined) averages about 8 Mbps (e.g. 480p at ~1.5 Mbps, 720p at ~2.5 Mbps, plus 1080p at ~4 Mbps).
  • Connection Handling: The load balancer maintains TCP/HTTP connections for streaming protocols (or even raw RTMP, if used) and may perform SSL/TLS termination.
  • Overhead: Always allow extra headroom for protocol overhead, bursty traffic, and connection management.

For 10 Concurrent Egress Streams

  • CPU:
    • 6–82 cores

    • ThisFor givesbasic headroompacket toforwarding runand roughlyconnection 20 concurrent encoding taskshandling (2even perwith stream)SSL plustermination), managea themodern passthrough.dual‑core CPU should suffice.

  • RAM:
    • 4 GB minimum (8 GB recommended)

    • To handle OS buffers, connection tracking, and any light processing.

  • Network:
    • 1 Gbps NIC

    • Estimated Egress Throughput: 10 × 8 Mbps ≈ 80 Mbps
    • A 1 Gbps link provides ample headroom.

For 100 Concurrent Egress Streams

  • CPU:
    • 4–8 cores

    • More cores help manage a higher number of concurrent connections, especially if SSL termination or deeper packet inspection is involved.

  • RAM:
    • 8 GB minimum (16 GB recommended)

    • EnsuresAdditional smoothmemory operationis withuseful multiplefor encodera processeslarger connection table and OS overhead.buffering.

  • Network:
    • 1 Gbps NIC might be borderline

    • Estimated Ingress:Egress Throughput: 10100 × 4 8 Mbps ≈ 40 800 Mbps
    • EstimatedRecommendation: Egress:Use 10a ×10 Gbps (1.5NIC +or 2.5 + 4) ≈ 80 Mbps
    • Thedual 1 Gbps linkNICs easily(bonded) coversto thiscomfortably withhandle roombursts forand spikes.protocol overhead.

For 1001,000 Concurrent Egress Streams

Important:
Handling 1,000 concurrent streams (totaling around 8 Gbps egress) is a high-demand scenario. Often, a single load balancer at this scale is implemented as part of a distributed architecture with clustering or specialized hardware.

  • CPU:
    • 16–32 cores

    • EachA stream produces 2 CPU-intensive transcoding jobs; a largerhigh core count is critical.needed Theto highermanage rangethousands isof advisedsimultaneous forconnections pureand softwarepotential encoding.SSL termination or other processing tasks.

  • RAM:
    • 16 GB minimum (32 GB recommended)

    • MoreTo concurrentefficiently encodingmanage tasksa willlarge benefitconnection fromtable extraand memory.OS/network buffers.

  • Network:
    • 10 GbpsHigh‑throughput NIC (or aggregated connections)NIC(s):

    • Estimated Ingress:Egress Throughput: 1001,000 × 4 Mbps ≈ 400 Mbps
    • Estimated Egress: 100 × ~8 Mbps ≈ 800 Mbps8 Gbps
    • AlthoughRecommendation: calculatedAt bitrates are below 1 Gbps, usingleast a 10 Gbps NIC, providesthough ampledepending headroomon forburstiness and protocol overhead, bursty traffic, and potential increases in bitrate if you choosemight higher-qualityconsider settings.aggregated NICs or a 40 Gbps solution to ensure stability.

Additional Considerations

  • HardwareDistributed Acceleration:Architecture:
    IfFor you1,000+ canstreams, useconsider GPUsdeploying multiple load balancers behind a front‑end DNS or dedicated encoding hardware (NVENCload akabalancing NVIDIA),solution youto canspread significantlythe reduce CPU requirements. For large-scale transcoding, this is often a more cost‑effectivetraffic and energy‑efficientprovide approach.redundancy.

  • Scalability:
    For very high concurrency (hundreds to thousands of streams), consider a multi‑server or cloud‑based transcoding farm that distributes the load rather than relying on a single box.

  • Encoding SettingsMonitoring & Quality:
    More aggressive encoding quality settings will increase CPU load. Tailor these recommendations based on your specific quality versus resource trade‑offs.

  • Redundancy & Future Growth:Scalability:
    Always plan withfor some extra headroom forto unexpectedhandle spikestraffic spikes, and futuremonitor scalingyour needs.system closely to adjust resource allocation as needed.

These guidelines provide a starting point to help you size your hardware. Actual requirements can vary significantly depending on your exact situation.

Back to top