Skip to content

Latest commit

 

History

History
27 lines (16 loc) · 1.24 KB

README.md

File metadata and controls

27 lines (16 loc) · 1.24 KB

gigabit-io-bench

An attempt to identify poor languages that cannot deal with gigabit speeds of disk, network, video or other type of IO.

Slow languages has always had an excuse: they weren't as fast as C, but they were still capable to saturate IO devices.

Here we'll try to identify and put shame on languages that cannot saturate a gigabit link using simple workloads that ought to be IO bound.

udp-fixed-localhost-flood Benchmark

Send as many UDP datagrams to 127.0.0.1:9999 as possible. Each datagram should carry 64 Q letters.

Hardcode the number of datagrams that makes the benchmark last about 30 seconds on your machine. Output the hardcoded calibration constant (number of datagrams).

1M pps udp @ 64 bytes = 1M * (58 + 64) * 8 = 976 Mbps

Language Packets Time pps Gbps
Node.JS 1M 30s 0.03M 0.03
Perl 10M 30s 0.3M 0.3
Haskell 100M 40s 3M 3

file-fixed-sequential-write Benchmark

Write as many data as possible to bench.bin file as possible. 8 MB blocks of Q letters are acceptable.

Hardcode amount of data that makes the benchmark last about 30 seconds on your machine. Output the hardcoded calibration constant (final file size in bytes).