> Am 19.04.2015 um 15:16 schrieb jb <justinbeech@gmail.com>:
>
> At least in my experience unless your most used static files exceed in size your available RAM, or are changing, they are effectively cached by the OS anyway.
>


Normally, yes.
Hence the reason why phk wrote Varnish, when he saw what squid was (and still is) doing...
But is that the case with NFS, too?
I thought there was some caching, too. But I’m not sure.


> So storing them on a ram disk is really doing the same or worse job than just letting the OS store them and serve them from its file cache memory pages. Plus the OS has the advantage of knowing which are less frequently used and can be purged.


Yep, that’s why I was asking.

If his data-set was _very_ big (in the large multi-TB region) and he had a couple of small SSDs to cache stuff, while at the same time the size of SSDs was about the size of the most requested files, it /could/ make sense.
But OTOH, you could also just install FreeBSD and use the SSDs as L2ARC and let the OS do the rest ;-)
Even the usefulness of L2ARC is often questioned by people familiar with the matter…

OS caching is _very_ hard to beat.
_______________________________________________
nginx mailing list
nginx@nginx.org
http://mailman.nginx.org/mailman/listinfo/nginx