Take 1 TCP listen socket and duplicate it across N threads so you get some sweet multi-thread action happening.
Checkout the speed up for PearDB:
Before
bench λ wrk -d 10 -c 24 -t 24 --latency -s readonly-pipelining.lua http://127.0.0.1:8888 ~/d/p/bench2
Running 10s test @ http://127.0.0.1:8888
24 threads and 24 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 2.41ms 811.15us 8.62ms 86.40%
Req/Sec 7.14k 1.52k 10.44k 68.91%
Latency Distribution
50% 2.09ms
75% 2.40ms
90% 3.95ms
99% 5.33ms
1633294 requests in 9.97s, 205.30MB read
Non-2xx or 3xx responses: 1469582
Requests/sec: 163850.16
Transfer/sec: 20.60MBAfter
bench [master] λ wrk -d 10 -c 24 -t 24 --latency -s readonly-pipelining.lua http://127.0.0.1:8888 ~/d/p/bench2
Running 10s test @ http://127.0.0.1:8888
24 threads and 24 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 1.54ms 2.49ms 48.06ms 95.51%
Req/Sec 14.52k 8.54k 51.78k 68.47%
Latency Distribution
50% 1.06ms
75% 1.78ms
90% 2.94ms
99% 7.90ms
3272402 requests in 9.96s, 411.25MB read
Non-2xx or 3xx responses: 2945770
Requests/sec: 328413.92
Transfer/sec: 41.27MB