I became curious how fast fdatasync (a better version of fsync) is, and asked LLM to generate a microbenchmark for me. The benchamrk accepts records size, number of records, opens a file, and appends records one by one, making fdatasync after each one. On my SSD, the throughput is 120 fsyncs/second, or ~0.5 MB/s. So if I had Redis to flush the log after each operation, it wouldn't be able to serve those thousands of connections. That's why Redis is either used without AOF, or they flush the data once per second. And that's why SQL databases might look "slow".

Of course one could optimize this - for example, while fsync is being executed, we could accept the queries from other clients and execute them, and once previous flush finishes, flush multiple transactions at once. However, I am not sure if Redis can do this due to being single-thread.

And obviously maybe there are problems with drivers, or with my consumer-level SSD and maybe "professional" SSDs can do more flushes per second.

Writing the code took less than a minute. I often do microbenchmarks now because it is so easy.