Insertion Benchmarks for TypeScript

[ibilux]

Hello everyone,

In the Insertion Benchmarks, where the graph shows the total time to insert 100k records, I noticed that C# and Rust took around 4 seconds, while TypeScript took over 16 seconds (about 4 times slower).

Can anyone explain why this might be the case?

[ignatz]

Can anyone explain why this might be the case?

Quoting from the benchmark page:

Comparing the Dart clients, both median read and insert latencies were about 5x lower for TrailBase. Curiously, TrailBase is fast enough that we’re being bottlenecked by the client runtime. The minimum latency you can expect from Dart and Node.js is around 3-5ms.

There are probably multiple factors at play:

• It's important to note that the logic in each language is rather trivial so what you're seeing is probably dominated by fixed-cost communication overhead.
• Even then, it's not far off what we'd expect from other language benchmarks. C# and Rust will typically perform 5x to 10x better. Rust is am AoT compiled language, with many static optimizations and w/o GC runtime overhead (in line with C/C++, Zig, ...), which typically performs best.
• Also Node executes concurrently on an event loop, while C# and Rust are truly parallel.
• Benchmarks never measure how fast something can go, they measure "how fast someone managed to make it go". I'm sure if you spawned more isolates and minimized allocations you could eek out a bit more performance from node.