(30th Sep, 2013 04:12 AM)IriDium Wrote: [ -> ]The only Xbian difference I can really see is the swap file is smaller (Guess because of the amount of available storage).
It's not a definitive test as I cannot guarantee they are both identical but I would have thought there shouldn't be much difference. I was hoping a) would have been faster, but it appears that maybe a USB installation is quicker.
Any ideas why? Is the I/O on the SD card slower than the USB ?
partswap parameter (creating a partition at end of the card) is not considering actual device size and is fixed to 256MB. if on b) is total swap bigger, then probably because both SD and USB contain swap partitions?
until actually the swap would be full, there is can be no difference in core speed (0% utilization of 1gb swap is no different to 0% util on 256mb swap).
just theoretically (without touching the setups myself) the a) must boot faster as beta2 will skip initramfs phase if root device is located on mmcblk0. this makes 6 seconds (this skip will be active after one reboot into the same media - so you put a) starts, reboot and initramfs will not be loaded until root= parameter is outside mmcblk0. whether initramfs.gz is used for next reboot is decided during reboot / shutdown and setting is local to the SD card being active (just for info)).
then important role in your test is how and when you created the filesystems (whether flash Beta1 + update, or one was copy via xbian-config etc). also, Beta1 img filesystem was created with older structure (older kernels) and with different parameters - it uses leaf size of 4K and duplicate metadata profile (each metadata is written twice). Beta2 img will be crated (and copies via xbian-config are) with 16K leaf and single profile. The perceived latency and speed of metadata operations are miles faster.
I'm running SD class10 16gb A-DATA card (I reformated it already, so "new" standard). Now created xbian-copy copy to micro class4, size of 8GB. to be honest I was expecting difference due to 10/4class, but the class10 loads xbmc in 31s, the micro class4 in 29s (standard XBian clocking, timing taken from /run/upstart-ev.log as time when xbmc-loaded service is started - what is triggered by XBMC itself as last event after all internal services are started, all python threads started, addons with autostart started and skin is fully loaded including UI objects active). second reboot measured (so no-initramfs) and install is clean Beta1.0 image + update to Beta2. nothing else. eth0 used and set to DHCP mode no non-Beta2 switches or tuning, nor disabled services. With clocking to 930/420/500/2 and static IP, it is up in 23s.
I assume the class10 seems to be slower because of cca 100 snapshots, while the class4 was just clean like a virgin.
how you measure or got the feeling about one being slower / faster ?
btw: just to give you some idea how actual setup affects for instance boot, AMBER with large bitmaps, +6s, 1channel plugin installed +2-3s. Wifi plugged in with working configuration +3s (with parallel eth0), other SD card/USB stick plugged in +3s, USB hub with NO DEVICES attached +1-5s (depends on number of ports to scan) etc.
So when I want to compare or check speed between changes, I always do a NEW snapshot from @safe (Beta1.0), reboot to the new snapshot install updates, reboot, reboot. Then I get comparable numbers.
But it is quite interesting topic indeed specially on the poor weak RPI cause even how you are breathing will change factors and timing in -5+5s
(30th Sep, 2013 10:48 AM)darrylb Wrote: [ -> ]Ok... So, I was able to (from a clean image) actually get Beta 2 installed. Now, I just run into the problem that the RaLink RT5370 (http://www.ebay.com/itm/171027110556?ssPageName=STRK:MEWNX:IT&_trksid=p3984.m1497.l2649)
is it this one problem?
https://github.com/xbianonpi/xbian/issues/421
how you got it working on previous releases? if you share info, no problem to implement.
(30th Sep, 2013 08:48 AM)wind-rider Wrote: [ -> ]Also - the layout of the fstab was a bit different than I was used to (it used LABEL=, @root etc), and the guides I read didn't cover changing that. That's also why I asked.
I promised - I will deliver - anyhow don't bother for rootfs "/", once mounted it can be missing from /etc/fstab, /home, /lib/modules will have to go out as well (cause no subvolumes on NFS) and I have to check all upstart dependencies for possible mounted or mounting events (they were used before, but plan was to remove them to make Beta2 as less dependent on actual boot device as possible (even on filesystem type).
just out from interest:
https://github.com/xbianonpi/xbian-package-initramfs/pull/5
6 months old. Nobody was using beside me and I stopped when EXT4->BTRFS change to be on the same system as users. Since then lot happened and NFS was never tested again. I go on it now, so +-1h.
(30th Sep, 2013 02:36 AM)Airwolfuk Wrote: [ -> ]My versions were from last week (Monday). I am running the upgrade now and will report back tomorrow.
ok, thanks
) and they are both configured roughly the same (same skin and add-ons) both on the same RPi and peripherals.