[dspillett]

Only if memory gets low, otherwise it'll stay in RAM and give the benefit GGP intended. IIRC tmpfs data shouldn't be evicted to swap just to allow more room for cache, or if an app requests a large chunk of memory but doesn't use it, just to allow more room for application pages that are actively in use.

Normal case: tmpfs data stays in RAM

Worst case: it is pushed to swap partitions/files, which is no worse than it being in a filesystem on physical media to start with (depending on access patters and how swap space is arranged it may still be a little more efficient).

It isn't quite the same as /tmp being on disk anyway but under normal loads in cache, because the data will usually get written to disk even if only ever read from cache and the cached data from disk will be evicted to make room for caching other data where tmpfs data is less likely to.

[anyfoo]

True overall, but I think it makes a lot of sense to evict rarely used data in tmpfs to swap, so that the DRAM it occupied can be used for valuable caches instead of holding some obscure temporary data that will be rarely or ever accessed.

Rarely used data that got evicted then behaves more or less like a normal /tmp filesystem when it does eventually get accessed, i.e. it gets read in from disk, while other data still gets all the benefits from tmpfs (e.g. ephemerality).

(If you take the thought experiment to its logical conclusion, you'll anyway end up in transparent hierarchical storage a la AS/400, where all data is just addressed by a single pointer in a very very large address space and the OS decides where that currently points to, but let's stay within the confines of what we're mostly used to...)

[cowsandmilk]

Use of /tmp on regular file system has almost the same behavior because the kernel has a file system cache… if you’re using the file, it will remain available in RAM. There’s some subtle differences, but I’ve seen enough benchmarks around this to have realized that tmpfs doesn’t really have an impact.

[anyfoo]

Yeah, that's why I think the prime feature of tmpfs is more ephemerality than anything else.