José David Baena – Software Engineer https://josedavidbaena.com Personal website and blog where I document thoughts, ideas, and interests in software engineering, web performance, and open source technologies. en-us josedab@gmail.com (José David Baena) josedab@gmail.com (José David Baena) Fri, 01 May 2026 00:00:00 GMT https://josedavidbaena.com/blog/zero-copy/08-zero-copy-failure-cost-model Zero-Copy Loses on Small Payloads, NUMA Misses, and Page Pinning https://josedavidbaena.com/blog/zero-copy/08-zero-copy-failure-cost-model A derivable cost model for when zero-copy loses: small payloads, remote NUMA, pinned pages, fallback rates, and completion lag move the frontier. Fri, 01 May 2026 00:00:00 GMT josedab@gmail.com (José David Baena) zero-copyperformancelinuxbenchmarkingsystems https://josedavidbaena.com/blog/zero-copy/07-zero-copy-serialization-borrowed-memory Zero-Copy Serialization Borrows Memory Instead of Owning It https://josedavidbaena.com/blog/zero-copy/07-zero-copy-serialization-borrowed-memory Zero-copy serialization wins only while bytes stay alive, validated, aligned, and unchanged. Compare Arrow, FlatBuffers, Cap'n Proto, and Protobuf. Fri, 24 Apr 2026 00:00:00 GMT josedab@gmail.com (José David Baena) zero-copyserializationperformancememorysystems https://josedavidbaena.com/blog/zero-copy/06-rdma-gpudirect-memory-topology RDMA and GPUDirect Move the Bottleneck to Memory Topology https://josedavidbaena.com/blog/zero-copy/06-rdma-gpudirect-memory-topology RDMA and GPUDirect remove specific host copies, but real wins depend on registered memory, topology, congestion control, and observable fallbacks. Fri, 17 Apr 2026 00:00:00 GMT josedab@gmail.com (José David Baena) zero-copyrdmagpudirectperformancelinux https://josedavidbaena.com/blog/zero-copy/05-zero-copy-storage-direct-io-spdk Zero-Copy Storage Is More Than O_DIRECT https://josedavidbaena.com/blog/zero-copy/05-zero-copy-storage-direct-io-spdk Separate copy avoidance, cache bypass, host-DRAM touch, and CPU work across mmap, O_DIRECT, io_uring, ublk, SPDK, and NVMe paths. Fri, 10 Apr 2026 00:00:00 GMT josedab@gmail.com (José David Baena) zero-copystoragelinuxio-uringspdk https://josedavidbaena.com/blog/zero-copy/04-af-xdp-dpdk-kernel-boundary AF_XDP and DPDK Win by Moving the Kernel Boundary https://josedavidbaena.com/blog/zero-copy/04-af-xdp-dpdk-kernel-boundary AF_XDP and DPDK move the kernel boundary to different places. Choose by queue ownership, transport semantics, NUMA locality, and rollback cost. Fri, 03 Apr 2026 00:00:00 GMT josedab@gmail.com (José David Baena) zero-copynetworkinglinuxperformancedpdk https://josedavidbaena.com/blog/zero-copy/03-io-uring-zero-copy-receive io_uring Zero-Copy Receive Trades Copies for Buffer Ownership https://josedavidbaena.com/blog/zero-copy/03-io-uring-zero-copy-receive Stable io_uring receive cuts syscall and buffer overhead. ZCRX can remove the copy, but only by turning RX memory into a buffer-ownership contract. Fri, 27 Mar 2026 00:00:00 GMT josedab@gmail.com (José David Baena) io-uringlinuxzero-copynetworkingperformance https://josedavidbaena.com/blog/zero-copy/02-tls-kafka-hidden-tradeoff Kafka TLS: Where the JVM Zero-Copy Fetch Path Ends https://josedavidbaena.com/blog/zero-copy/02-tls-kafka-hidden-tradeoff Trace Kafka's file-backed Fetch payload through OpenJDK, JSSE, kTLS, and NIC offload without confusing their copy boundaries. Fri, 20 Mar 2026 00:00:00 GMT josedab@gmail.com (José David Baena) kafkaperformancetlslinuxzero-copy https://josedavidbaena.com/blog/zero-copy/01-philosophy-not-technique Zero-Copy Starts by Removing Work That Adds No Value https://josedavidbaena.com/blog/zero-copy/01-philosophy-not-technique Trace each byte owner, separate Linux transfer mechanisms, and decide when copy avoidance earns its lifetime, fallback, and observability costs. Fri, 13 Mar 2026 00:00:00 GMT josedab@gmail.com (José David Baena) zero-copyperformancesystemslinuxnetworking