Canonical Announces ETrace As New Linux Application Tracing For Performance/Debugging

[phoronix]

Phoronix: Canonical Announces ETrace As New Linux Application Tracing For Performance/Debugging

Canonical has announced ETrace as a new application tracing tool designed for debugging and performance profiling of Snap packages but can also be used with any Linux binary applications...

http://www.phoronix.com/scan.php?page=news_item&px=Canonical-Ubuntu-ETrace

[Jumbotron]

Well....GOOD ON YA' Canonical. With ETrace and the new compression algorithm being used for Snaps I think we'll see very significant performance uplifts very soon for Snaps which of course desperately need them.

I for one, on the whole, like Snaps more than Flatpaks. They almost to an app, seem more performant (once you get over the obvious boot up time lag vs Flatpaks) than Flatpaks. Particularly in some games, I've seen 20-50% slower frame rates in Flatpaks than the same game in Snaps. Also...last year during a weird glitch that was preventing me from installing LibreOffice from their site, I tried to load the Flatpak version from Flathub. Failure also. I tried installing other Flatpaks to see if my Flatpak setup plus Flathub was at fault. Nope..other Flatpaks installed and ran just fine. Then I went to the Ubuntu App store, downloaded the Snap. Perfect. No issues other than the long initial load times. Afterwards startup was very quick. And no operational issues.

Snap...for all its obvious weaknesses has never failed me. And now it looks like Canonical is seriously getting on the ball concerning performance issues particularly on the startup.

[Britoid]

Well....GOOD ON YA' Canonical. With ETrace and the new compression algorithm being used for Snaps I think we'll see very significant performance uplifts very soon for Snaps which of course desperately need them.

I for one, on the whole, like Snaps more than Flatpaks. They almost to an app, seem more performant (once you get over the obvious boot up time lag vs Flatpaks) than Flatpaks. Particularly in some games, I've seen 20-50% slower frame rates in Flatpaks than the same game in Snaps. Also...last year during a weird glitch that was preventing me from installing LibreOffice from their site, I tried to load the Flatpak version from Flathub. Failure also. I tried installing other Flatpaks to see if my Flatpak setup plus Flathub was at fault. Nope..other Flatpaks installed and ran just fine. Then I went to the Ubuntu App store, downloaded the Snap. Perfect. No issues other than the long initial load times. Afterwards startup was very quick. And no operational issues.

Snap...for all its obvious weaknesses has never failed me. And now it looks like Canonical is seriously getting on the ball concerning performance issues particularly on the startup.

Eh, Flatpaks are just bind mounts with clever namespacing, Flatpak itself will have zero affect on frame rates. It's likely caused by a different version of mesa in the freedesktop runtime. This is why Flatpak has very little start time penalty.

Meanwhile, Snap will lower framerates because it introduces I/O overhead as a result of it's loop system. The kernel has a natural I/O overhead here and there's not much around it. That's also why Flatpak will always be faster than Snap at start up times, Flatpak is just bind mounts which have no overhead and are created nearly instantly, whilst there's a read penalty for loopback filesystems.

[andre30correia]

Eh, Flatpaks are just bind mounts with clever namespacing, Flatpak itself will have zero affect on frame rates. It's likely caused by a different version of mesa in the freedesktop runtime. This is why Flatpak has very little start time penalty.

Meanwhile, Snap will lower framerates because it introduces I/O overhead as a result of it's loop system. The kernel has a natural I/O overhead here and there's not much around it. That's also why Flatpak will always be faster than Snap at start up times, Flatpak is just bind mounts which have no overhead and are created nearly instantly, whilst there's a read penalty for loopback filesystems.

not my experience and I use both 3d software works much better with snaps

[Jumbotron]

Eh, Flatpaks are just bind mounts with clever namespacing, Flatpak itself will have zero affect on frame rates. It's likely caused by a different version of mesa in the freedesktop runtime. This is why Flatpak has very little start time penalty.

Meanwhile, Snap will lower framerates because it introduces I/O overhead as a result of it's loop system. The kernel has a natural I/O overhead here and there's not much around it. That's also why Flatpak will always be faster than Snap at start up times, Flatpak is just bind mounts which have no overhead and are created nearly instantly, whilst there's a read penalty for loopback filesystems.

Wrong. Multiple Flatpak games vs Snap games. Almost to a game Flatpaks had lower framerates than Snap games. Period. Full stop.

Absolutely does not matter if it has to do with different versions of Mesa in the freedesktop runtime. If Flatpaks can't overcome that situation and Snaps can...well...Flatpaks are at fault for that. Period. Full Stop.

[Jumbotron]

not my experience and I use both 3d software works much better with snaps

I have found that as well.

[lyamc]

I have had less issues with snaps than flatpaks.

[edwaleni]

Some privacy folk are disturbed at the ability of Snaps to be updated at a root level without user authorization.

The snap for Google Chrome apparently does this with impunity.

If this is incorrect, please post.

[simonsaysthis]

I have had less issues with snaps than flatpaks.

I second that, especially for third-party apps. I didn't care for all the anti-Canonical propaganda and just looked for a solution that serves my needs.

[pal666]

If Flatpaks can't overcome that situation and Snaps can...well...Flatpaks are at fault for that..

if applicaation(flatpak) on your distro(ubuntu) works suboptimally, the only one to blame is distro vendor. not "pulseadudio breaks my sound" but "ubuntu breaks my sound"