blymn@baesystems.com.au (Brett Lymn)  wrote:
 > Doesn't the mkdir play with ufs metadata which, IIRC, needs to be
 > written _synchronously_ to disk which means that if your disk
 > subsystem is not fast you will lose on synchronous operations whereas
 > normal operations are using the buffer cache which _may_ mask a slow
 > disk subsystem somewhat (that was what it is for ....)

3500 I/O's to do a mkdir?