On 12/13/13 04:08, Jérôme Frgacic wrote: > Hi misc, > > I have installed OpenBSD on an USB stick (a Kingston DataTraveler G3). > Nevertheless, the system is quite slow... For example, I recently install > firefox or, more precisely, those packages: > ...[snip lots of packages]... > > which have a total size of 67Mo and it tooks ~17min to finish. If I compute > correctly, this give me an average of ~67ko/sec which, I think, is quite bad. > I precise that I download and store all the packages on an mfs partition, so > this is really the time consumed to install them. Also, all the partitions are > mounted with the softdep and the noatime options.
good start, definitely. Keep in mind, dividing total package size by time required produces a number with units of "bytes of packages installed per second", which is not overly interesting, and VERY different from "bytes written per second", as a lot of work goes into a package install beyond a single linear file write. > Is this normal ? Did I make any mistakes in the organisation of the system on > the stick ? Is there something I can do to improve the performance of the I/O > ? First of all... USB isn't fast, flash is not fast. Aaaannddd.... you have an nvidia based machine. > Below, you will find the output of the fdisk, disklabel, usbdevs and dmesg > commands. > > Kind regards, > > Jérôme > > ----------------------------------------------------------------------- > > # fdisk sd0 > Disk: sd0 geometry: 3762/255/63 [60437492 Sectors] > Offset: 0 Signature: 0xAA55 > Starting Ending LBA Info: > #: id C H S - C H S [ start: size ] > ------------------------------------------------------------------------------- > 0: 0B 0 1 2 - 522 43 33 [ 64: 8388608 ] Win95 > FAT-32 > 1: 00 0 0 0 - 0 0 0 [ 0: 0 ] unused > > 2: 00 0 0 0 - 0 0 0 [ 0: 0 ] unused > > *3: A6 523 0 1 - 3761 254 63 [ 8401995: 52034535 ] OpenBSD > you *MIGHT* find a small improvement if your OpenbSD partition is aligned to a 4k (8 sector) count. Would be interesting, I would bet it wouldn't be "human noticable" (i.e., at least 2x performance), but might be statistically significant. And I might be wrong -- it might be noticable. ... > # dmesg > OpenBSD 5.4 (GENERIC.MP) #44: Tue Jul 30 12:13:32 MDT 2013 > dera...@i386.openbsd.org:/usr/src/sys/arch/i386/compile/GENERIC.MP ...[snip. Thanks for providing... lots of nvidia hw] Would be interesting to try your test on a non-nvidia machine. Nick.
