Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Any decent SSD has capacitor (enterprise) or battery backed (phones) DRAM. Therefore, a sync write is just “copy the data to an I/O buffer over PCIe”.

For databases, where you do lots of small scattered writes, and lots of small overwrites to the tail of the log, modern SSDs coalesce writes in that buffer, greatly reducing write wear, and allowing the effective write bandwidth to exceed the media write bandwidth.

These schemes are much less expensive than optane.



I have tried multiple enterprise SSD's, for sync writes. Nothing comes close to Optane Dimm, even Optane NVMe is 10x slower than PDIMMS.

https://forums.servethehome.com/index.php?threads/so-i-teste...


Yeah, but what percentage of the time was the CPU stuck blocked on the memory bus for the two workloads?

Also, how was it guaranteeing the page writes were atomic?


You are correct that writes are not guaranteed to be atomic. PostgreSQL has a feature to handle torn pages: full_page_writes = on, which is enabled by default. This means you do double write.

For RocksDB I believe torn pages are not a problem as SST files are immutable. But correct me if I am wrong.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: