Hacker News Reader

Comment by magicalhippo | Hacker News Reader

> Huh? What are you talking about? ZFS has had a write cache from day 1: ZIL.

The ZIL is specifically not a cache. It can have cache-like behavior in certain cases, particularly when paired with fast SLOG devices, but that's incidental and not its purpose. Its purpose is to ensure integrity and consistency.

Specifically, the writes to disk are served from RAM[1], the ZIL is only read from in case of recovery from an unclean shutdown. ZFS won't store more writes in the SLOG than what it can store in RAM, unlike a write-back cache device (which ZFS does not support yet).

[1]: https://klarasystems.com/articles/what-makes-a-good-time-to-...

View on HN

That explains the results I've got last week when I was doing some tests on an slow external HDD and a SLOG file placed on my internal SSD, the write speed dropped much faster that I expected. I found it disappointing that ZFS needs to keep the data in RAM and on the SLOG dev. You really need a lot of RAM to make full use of it. And a larger SLOG dev than RAM space available makes no sense then? A real write cache would be awesome for ZFS.

> And a larger SLOG dev than RAM space available makes no sense then?

Exactly, and it's even worse. By default ZFS only stores about 5 seconds worth of writes in the ZIL, so if you have a NAS with say a 10Gbe link that's less than 10GB in the ZIL (and hence SLOG) at any time.

> A real write cache would be awesome for ZFS.

Agreed. It's a real shame the write-back cache code never got merged, I think it's one of the major weaknesses of ZFS today.

I haven't tried Nexenta in a long long time, but it seems like it was part of their allocation classes code [1]. The comment at the top of the file says the implementation there doesn't support WBC, but [2] seems to suggest otherwise...

So if you really wanted to, you could try looting code from there, though I believe Nexenta's implementation of that was either orthogonal or significantly different from the one that landed in OpenZFS...

[1] - https://github.com/Nexenta/illumos-nexenta/blob/release-5.3/...

[2] - https://github.com/Nexenta/illumos-nexenta/blob/release-5.3/...

That you for being right!

So many people misunderstand the ZIL SLOG, including the guy you are responding to.

I think most posts on the internet about the ZIL SLOG do not explain it correctly and thus a large number of people misunderstand it.

Yes it took me some time to fully grasp how the ZIL and SLOG worked and how it is not a cache.

There was some work done[1] on proper write-back cache for ZFS by Nexenta several years ago, but it seems it either stalled or there was a decision to keep it in-house as it hasn't made its way back to OpenZFS.

[1]: https://openzfs.org/wiki/Writeback_Cache

Reply on news.ycombinator.com