ProcessPool vs ThreadPool: GIL CPU Lab with Real Numbers
Hands-on ProcessPool vs ThreadPool lab: pure-Python threads lose to processes, hashlib threads help, and sleep I/O favors threads.
30 Sept 20265 min read
Topic
sre
111 posts
Loading posts…
epoll vs select Toy: Ready-Wait Latency and the FD_SETSIZE Cliff
Hands-on epoll vs select toy: ready-wait flat ~3.5 µs (8→900 FDs); select scales 4.6→16.9 µs then hits fd≥1024 cliff. Python numbers, no Docker.
30 Sept 20265 min read
sendfile vs Userspace Copy: Localhost TCP File Serve Lab
Hands-on sendfile vs read+sendall lab: 256 MiB p50 4.41 vs 2.34 GB/s (~1.9×); copy collapses at 4 KiB. Warm localhost TCP measured, no Docker.
30 Sept 20265 min read
mmap vs read (+ O_DIRECT): Localhost Sequential Scan Lab
Hands-on mmap vs buffered read lab: 256 MiB p50 11.5 vs 5.4 GB/s; O_DIRECT ~0.9 GB/s warm-cache bypass. Fake endpoint-only GB/s disclosed. No Docker.
30 Sept 20265 min read
Nginx Gzip On vs Off: Localhost Wire Size and RPS Lab
Hands-on nginx gzip on/off lab: HTML ~107x smaller at level 1; JSON RPS ~1987 off vs ~1031 l1 vs ~563 l6 on localhost. Measured numbers, no Docker.
30 Sept 20266 min read
Pipe vs Tmpfile IPC: Localhost Throughput and RTT Lab
Hands-on pipe vs tmpfile IPC lab: 5.32 vs 1.51 GB/s bulk (fsync 0.57); 64 B RTT p50 3.3 us vs 70 us vs 289 us. Localhost measured, no Docker.
30 Sept 20266 min read
What Is Nagle's Algorithm? TCP_NODELAY vs Nagle Tested
Hands-on TCP_NODELAY vs Nagle: 1-byte reuse 124k vs 73k RPS; 32×16B writes favor Nagle (~1.7× fewer OutSegs). Delayed-ACK mean trap on localhost.
30 Sept 20266 min read
ulimit Soft vs Hard File Descriptors: EMFILE Lab with Real Numbers
Hands-on RLIMIT_NOFILE lab: soft=64 → EMFILE after 58 held accepts; soft=512 takes 200/200. Soft raises without root; hard does not. Real numbers.
30 Sept 20265 min read
SO_REUSEPORT vs Single Listen: Independent Queues and ListenOverflows
Hands-on SO_REUSEPORT lab: ss shows 4 listen queues; backlog=8 flood 72% vs 62% OK; ~41% fewer ListenOverflows. Real localhost numbers, no Docker.
30 Sept 20266 min read
TCP Listen Backlog Lab: somaxconn, ListenOverflows, and Silent Drops
Hands-on TCP backlog lab: backlog=1 + slow accept → 34/200 OK, ListenOverflows +357; backlog=128 stays 200/200. Real ss + netstat numbers.
30 Sept 20266 min read
Unix Domain Socket vs TCP Localhost: Latency and Throughput Lab
Hands-on UDS vs TCP loopback echo: connect-per ~3× RPS (49k vs 16k); reuse ~1.3×. Real Python numbers, no Docker.
30 Sept 20266 min read
HTTP Keep-Alive vs Connection: close: Localhost Numbers and a Simulated-RTT Trap
30 Sept 20267 min read