A deep dive into Sequential Consistency, Total Store Order (TSO on x86-64), and weak memory models on ARM64. How hardware store buffers and cache coherency protocols dictate lock-free algorithm correctness.
How we bypassed the Linux kernel network stack overhead using XDP (eXpress Data Path) and UMEM memory-mapped ring buffers, achieving line-rate 10Gbps packet filtering with under 3% CPU utilization.
Practical lessons from diagnosing GC pacing pauses, lock contention on channel multiplexers, and using arena allocators and sync.Pool to reduce garbage collection pressure on latency-critical paths.
Handling duplicate network packets, retry storms, and clock drift in distributed consensus systems without relying on heavy distributed lock managers.
A cache-line padded, lock-free circular ring buffer implementing SPMC and MPMC semantics in Go.
Terminal UI tool for inspecting live eBPF map contents and packet counters attached to NIC queues.
Microsecond-precision RTT and packet loss measurement daemon for QUIC and HTTP/3 streams.