Date: Sun, 12 Jan 1997 12:34:40 -0700 (MST) From: Terry Lambert <terry@lambert.org> To: henrich@crh.cl.msu.edu (Charles Henrich) Cc: terry@lambert.org, freebsd-hackers@FreeBSD.org Subject: Re: mount -o async on a news servre Message-ID: <199701121934.MAA26007@phaeton.artisoft.com> In-Reply-To: <199701121905.OAA13998@crh.cl.msu.edu> from "Charles Henrich" at Jan 12, 97 02:05:53 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > You seem to be confusing "noatime" with "async". The "async" option > > acts pretty much as you describe. > > You've misread what I said. Using atime with sync's done every 300 > seconds I would beleive is very similar in performance to async, > without the risks. This is of course only in the case of a news spool. Ah... you are claiming the only metadata events for a read-only news spool are atime related. I agree with you, then... > > You could rephrase this as "Is there any safe way to run 'async'?", > > with the answer being "Yes, if you only read from, and never post > > directly to, the 'async' server". > > No No, perhaps I wasnt clear. I was never advocating running async, its a big > bad idea IMHO. I was attempting to show that running with noatime and sync's > pushed back to be infrequent (300 seconds) you get nearly the performance win > of async, without the risks.. The "noatime" means "noatime", not "async atime". I think changing the sync frequency for update will have no effect, unless theres a conflict with the free buffer high water mark. Even so, if that happened, then you are trading update time for blocking access to buffers for some users (assuming the reclaim high water mark is set correctly, which I believe it is). Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199701121934.MAA26007>