gwyn@brl-smoke.ARPA (Doug Gwyn ) (12/20/87)
In article <939DBLCU@CUNYVM> DBLCU@CUNYVM.CUNY.EDU writes: >I have a follow up question on using soft links. Is there a pronounced >performance cost by using such a link, or is it negligible? I have trouble imagining an application where the cost would be noticeably more than some other approach that accomplishes the same thing. (I phrased this strangely, to acknowledge that the target of a symlink might be a remote file system, for example.)
rwa@auvax.UUCP (Ross Alexander) (12/22/87)
In article <6892@brl-smoke.ARPA>, gwyn@brl-smoke.ARPA (Doug Gwyn ) writes: > In article <939DBLCU@CUNYVM> DBLCU@CUNYVM.CUNY.EDU writes: > >I have a follow up question on using soft links. Is there a pronounced > >performance cost by using such a link, or is it negligible? > I have trouble imagining an application where the cost would be > noticeably more than some other approach that accomplishes the same > thing. (I phrased this strangely, to acknowledge that the target > of a symlink might be a remote file system, for example.) One of our machines here (auvax) used to have 3 ra81's, which had three user-account filesystems named (strangely enough) /usr[123]. This worked quite nicely until the day came when I had to pull a drive offline to test a new (and subsequently discovered, extremely bug-prone) version of Ultrix. This was forced on me by a change in the size of the inodes, but I digress. Now my user community, in their infinite wisdom, had scads of code containing pathnames of the form '/usr1/twit/bar/foo', et c., et c., (despite considerable flaming on my part well before this problem arose). Said pathnames being hardwired in, and not a few of 'em, either, I really was puzzled for a while, at least several minutes, as to how I could fake this thing without breaking every user-designed programme in sight. So I ended up repartitioning the two remaining 'public' drives with /usr[ab] partitionas, with /usr{1,3} being mapped into /usrb via symlinks in /, and /usr{,2} mapped into /usra ditto. The upshot of this is almost every path in the system now has a symlink in it. Performance has not noticeably altered. So much for the symlink penalty ;-) Ross Alexander @ Athabasca University, alberta!auvax!rwa