Thread on socket 0 → accessing socket 0 memory → ~80 ns latency Thread on socket 0 → accessing socket 1 memory → ~140 ns latency (over 70% slower)
: Early Unix was rewritten in the C programming language, making it highly portable across different hardware types. Unix Systems For Modern Architectures.pdf
In the early 1990s, Unix was a mess—not of code, but of scaling. Traditional Unix kernels were designed for single-CPU, single-bus systems. The kernel was a giant "Big Kernel Lock" (BKL): only one process could execute kernel code at a time. Thread on socket 0 → accessing socket 0