Benchmark 的一些记录

[bxb100]

记录一下看到的各种 benchmark 指标的网站, 仓库, 和博客

• YCSB A common use of the tool is to benchmark multiple systems and compare them.

• yet-another-bench-script YABS - a simple bash script to estimate Linux server performance using fio, iperf3, & Geekbench

使用的工具

• https://github.com/hatoo/oha 压测
• https://github.com/flamegraph-rs/flamegraph 火焰图
• https://github.com/sharkdp/hyperfine A command-line benchmarking tool

[bxb100]

ClickHouse 主导的和其它数据库性能的 benchmark	repo	website
https://github.com/ClickHouse/ClickBench	https://benchmark.clickhouse.com/

TLDR: All Benchmarks Are ~Bastards~ Liars.

[bxb100]

SigNoz 主导的和 ES, Loki 的性能比较

repo	blog
https://github.com/SigNoz/logs-benchmark	https://signoz.io/blog/logs-performance-benchmark/

Conclusion

• For ingestion SigNoz is 2.5x faster than ELK and uses 50% less resources.
• Loki doesn’t perform well if you want to index and query high cardinality data.
• SigNoz is about 13 times faster than ELK for aggregation queries.
• Storage used by SigNoz for the same amount of logs is about half of what ELK uses.

[bxb100]

Lambda 各个语言冷启动, 对 serveless edge 场景有借鉴意义

twitter	website
Twitter	https://maxday.github.io/lambda-perf/

---

Rust 在边缘节点计算真有前途, 不过不知道 WASM 效能是不是一样

[bxb100]

由于 JDK 19 virtual thread 出现, 需要将 synchronized 替换成 ReentrantLock 的需求

source	detail
Twitter
pgjdbc

---

尽量用 synchronized 不用 wait, notify 应该没问题

[bxb100]

AWS Network Testing (not benchmark)

source: https://twitter.com/haoel/status/1654655067365179393?s=20

https://docs.google.com/spreadsheets/d/1ayC9SV1kgGjB_VHa8pAuyxjMIIZf5bEh1kiYDTo8j6I/edit#gid=0

---

RTT + Cwnd 控制网速

[bxb100]

LLM Speed Benchmark https://thefastest.ai/

This site provides reliable measurements for the performance of popular models.

[bxb100]

Axum (Rust) vs Stdlib (Go): Performance Benchmark in Kubernetes #205