Log10 Loadshare | 360p |

Problem: A server with 1 core gives ( \log(2)=0.3 ), but a server with 0.5 core (if fractional allowed) is invalid.
Solution: Floor capacity at 1. For health checks, exclude dead servers from weight calculation.

Log10 Loadshare is a load-balancing weight assignment strategy where the share of traffic directed to a given server or resource pool is proportional to the base-10 logarithm of that server’s capacity metric (e.g., CPU cores, memory, network bandwidth, or request-handling throughput). log10 loadshare

If a high-capacity server starts failing requests, you can temporarily reduce its weight by treating its effective capacity as lower. Because weights compress, other servers can absorb the load without cascading failure. Problem : A server with 1 core gives ( \log(2)=0

| Myth | Reality | | :--- | :--- | | "log10 loadshare hides outliers." | No—it preserves outlier order. The largest raw value still has the largest log value. It only compresses the visual distance. | | "It only works for request rates." | False. It works for any positive load metric: bytes/sec, active connections, queue length, CPU steal time. | | "Zero is problematic." | Solved by the +1 offset. A server with 0 load is beautifully represented as 0. | | "It adds computational overhead." | Negligible. log10 is a cheap floating-point operation. Even at 1M requests/sec, the overhead is microseconds. | | Myth | Reality | | :--- |

If all servers have nearly identical capacity (e.g., all 8 cores), Log10 produces near-equal weights. That’s fine—but you lose no advantage. Round robin is equivalent.