Hi I have tested several Plasma isos since alpha1 until Beta0 and the last Beta1 round3 Each time I created a new pendrive... and used the same computer Several tests with or without persistent partition : The last tests with the last Beta1 are quite perfect without persistent partition 1) Without persistent partition the last Beta1 works quite perfectly with nouveau or Nvidia driver (since Beta0w or Beta0x from Giuseppe evrything is quite OK) We need 5 minutes until Plasma appears (the finish-install step allows to configure correctly the system before we can use Plasma) Launching a program displays it quite immediately (Firefox as an example is quite as quick as on my installed Mageia9) 2) With a persistent partition - the first boot for Nvidia takes 20 minutes before Plasma starts - the first boot for nouveau takes 5 minutes But Plasma or the use of any program is dramatically slowed down (the flash of the pendrive is ceaseless flashing - the second boot for Nvidia takes 4 minutes before Plasma starts - the second boot for nouveau takes 3 minutes before Plasma starts But for both of them Plasma and any program are dramatically slowed down, like after the first boot... As an example Firefox needs more than one minute to launch, and more than one minute to display the content of the two default tabs There something wrong about the presence of a persistent partition that seems to be used instead of the RAM of the computer (without this partition everything is fast for the pendrive) : Is it the use of the /temp/ folder from the /work/ folder of the persistent partition that slows down everything (alternating read and write implies that the led flashes incessantly) ?
See Also: (none) => https://bugs.mageia.org/show_bug.cgi?id=35020
IMHO it's more a "feature", due to poor performance of the particular USB pendrive on managing random RW access with small block transfers. See https://bugs.mageia.org/show_bug.cgi?id=35020#c231.
CC: (none) => ghibomgx
(In reply to Giuseppe Ghibò from comment #1) > IMHO it's more a "feature", due to poor performance of the particular USB > pendrive on managing random RW access with small block transfers. > > See https://bugs.mageia.org/show_bug.cgi?id=35020#c231. This might need a warning in the wiki about the use of a persistent partition when creating a pendrive with a live iso You need not only USB 3.1 port and pendrive, but even an high speed pendrive !!! Indeed most of the pendrives have bad writing performances for small blocks transfers : https://ssd-tester.fr/cle_usb_test.php Nota Bene : in the table the writing speed concerns only big files... for each pendrive the green test button shows a complete test revealing bad speed for writing small blocks in most of cases
(In reply to Philippe Didier from comment #2) > You need not only USB 3.1 port and pendrive, but even an high speed > pendrive !!! In my experience USB2 works well - it is much more important to use a decent pendrive, than using USB3+. And look more to speed of random writes, than the top speed on the pendrive package. High speed large file read is of course iportant for loading kernel and other big stuff, but some desktop environments (Plasma) seem to do lots of access, and system updating really need high random access speed, web surfing etc (caching i presume) also use much read. > Indeed most of the pendrives have bad writing performances for small blocks > transfers : > https://ssd-tester.fr/cle_usb_test.php > Nota Bene : in the table the writing speed concerns only big files... There are huge differenced between different sticks, and unfortunately random access writing speed is never advertised on the USB stick packages... I initiated testing some years ago, please anyone interested add their results, testing in the same or similar way for easy comparison: https://wiki.mageia.org/en/Storage_speed_test_results Also see https://wiki.mageia.org/en/Talk:Persistent_live_systems For ideas on how to improve for example speed of Firefox by forcing it to use RAM.
CC: (none) => fri
Hi Morgan I have explained that the table https://ssd-tester.fr/cle_usb_test.php gives only write and read speed for large files But you may find a complete test for each of the pendrive showing its speed for writing or reading small blocks : it's a pity when you see more than 90% of them have a write speed for small blocks less than 1 MB/sec...
Yes that is an example of a bad test site. Here is a better one, you can even sort it on 4k Write performance: https://usb.userbenchmark.com/ Also, anyone is welcome to add to the table I started in our wiki, where one of the test is exercising the package manager. We already explain importance of speed at https://wiki.mageia.org/en/Persistent_live_systems#Choosing_storage
(In reply to Philippe Didier from comment #4) > the table https://ssd-tester.fr/cle_usb_test.php gives > only write and read speed for large files I now see that a link from that page gives https://ssd-tester.com/top_usb_sticks.php which show 4k read and writes I have now added links at https://wiki.mageia.org/en/Testing_storage_speed#Other_test_sites
Hi Morgan Thanks for your advice Maybe you might too add some links inside https://wiki.mageia.org/en/Persistent_live_systems in its "Choosing storage" part for example https://usb.userbenchmark.com/ and https://ssd-tester.com/top_usb_sticks.php
Going back to comment 0: Philppe's results 'with a persistant partition' are apalling, and totally unuseable. "There something wrong about the presence of a persistent partition that seems to be used instead of the RAM of the computer" It is not clear whether the mere presence of a persistant partition matters if the system is run from the Live partition, or whether the problem happens if the system is installed in & run from the persistent partition. Maybe all this is nonsense - beyond my experience. (Away, so not subscribing to this bug).
At bottom of https://wiki.mageia.org/en/Persistent_live_systems#Choosing_storage now added link to links. (having only one set of links make it easier to maintain consistency when we find and add more there) Closing as Worksforme, as it is up to user to choose a decent stick and instructions are in wiki. Possibly we should note it in release notes among descriptions of Live systems, linking to https://wiki.mageia.org/en/Persistent_live_systems#Choosing_storage ? @josé what do you think?
Resolution: (none) => WORKSFORMECC: (none) => j.alberto.vcStatus: NEW => RESOLVEDSummary: Persistent partition slows down Mageia10 iso => Slow USB stick slows down Mageia Live when using Persistent partition (in wiki)Flags: (none) => in_release_notes10?, in_wiki+
Would be a good recommendation, need to read all the comments But a USB2 or USB3 memory will be good, but consider that more newer devices also limit the type of hardware where you can use it. I Let this in your hands ;)
CC: j.alberto.vc => (none)
(In reply to Lewis Smith from comment #8) > Going back to comment 0: > Philppe's results 'with a persistant partition' are apalling, and totally > unuseable. > "There something wrong about the presence of a persistent partition that > seems to be used instead of the RAM of the computer" > It is not clear whether the mere presence of a persistant partition matters > if the system is run from the Live partition, or whether the problem happens > if the system is installed in & run from the persistent partition. > Maybe all this is nonsense - beyond my experience. > > (Away, so not subscribing to this bug). Hi Lewis Sorry for my bad hypotheses in comment0 Fortunately Giuseppe enlightened me : testing my pendrives with kdiskmark showed that indeed their performances are apalling and caused the problems I encountered with a persistent partition (writing or reading or modifying little files is horribly slow : 0.4 MB/s on these pendrives !!!) I couldn't imagine that building Nvidia driver might need the use of a persistent partition if it is present !!!