I first struggled with my single-user Friendica instance after I added all followed accounts from Mastodon. Very slow, sometimes even not responding at all.
After tweaking some settings the site is running quite smootly. So it depends on the setup.
Yes, Mastodon out-of-the-box is faster on most machines, but Friendica can be made very usable.
Via clients like WhaleBird or Fedilab, I don’t feel any difference. @serenity
PHP memory limit up to 2048
Number of workers to 20
The workers now run every 5 minutes instead of 10 (cron)
Increased the values of pm.max_children , pm.start_servers, pm.min_spare_servers and
pm.max_spare_servers.
I’m running this setup on Yunohost in a VM with 2 cores and a total of 16 GB RAM. Currently only 7 or 8 GB are used, that’s including Mastodon on the same server.
Giving the friendica instance more resources really helped, the standard settings are too restrictive. @serenity@rom
@serenity@f00fc7c8@rom It’s not even sorta fast, but I’ve found it to be the most functional for me out of the bunch. In addition to Friendica, I’ve played around with Diaspora, Mastadon, Pleroma, Misskey, plus the forks Calckey and Akkoma. Friendica’s ability to easily follow tags and rss feeds, plus the lemmy and diaspora integration, as well as the ease of import/export really shows where some of the others are lacking. That said, UI/UX customization in Friendica leaves quite a bit to be desired.
@eshep@serenity@rom That’s not always because of a speed issue though. I usually see that on messages that were boosted or searched recently, in which case it’s no different from other AP servers.
@serenity@f00fc7c8@rom Good to know. I’ve noticed it mostly on post pulled in from hashtag follows. I thought maybe it could be something to do with how those are polled.
@eshep@serenity@f00fc7c8@rom If you’re following a hashtag you’ll see posts from people nobody on your instance is following and those won’t come in as they’re posted, but will be fetched as a side-effect of actions by people that are being followed.
My issue with Friendica is that it takes several seconds to render the “timeline” sometimes. Tempted to move the docker containers from the RPi that’s currently hosting them to my M1-pro MacBook.
@serenity@rom@eshep@f00fc7c8 In my experience it is not because Friendica is slow. then Its because someone I follow just commented a post and then the server had to download the post whereas before there wasnt a reason to because i didnt follow the OP.
@serenity@rom@eshep@f00fc7c8 I tried to import/move my profile to my own instance but I got an error saying page not found :S did you succeed with that`?
I agree with what you say. Is your experience that Friendica is slower than the others? My guess would be that Friendica is programmed in PHP which executes slower than other programming languages.
Friendica can be rather resource-intensive, especially for public instance. Some of them feel slow because they are somewhat overloaded. On badly overloaded servers delivery times suffer.
Also mostly static interface of Friendica which I like also means most operations reload the entire page. E.g. when you go to someone’s profile it can take the same time (say 8 seconds) on Friendica and Misskey/Pleroma/Mastodon but on the latter you don’t see the page redrawn - it stays the same and something spins loading the data. On Friendica you see full page reload. It is perceived slower.
Also in some aspects Friendica is inherently slower by design. E.g. when posting comment on Mastodon it is fired off right away (well, almost), on Friendica it is queued for delivery and the worker runs on interval (usually every 2-5 mins). It isn’t a problem but feels more like email or old school forum while other platforms can feel more like instant messaging.
@eshep@serenity@f00fc7c8@rom@anders Interesting, I didn’t know about daemon. Can you share how you have it configured? I set my server some years ago, I think it didn’t have this option back then but I’d like to try.
I assume that you don’t run it through Docker since you mentioned Cronjobs. Let me know if my assumption is wrong because then I can give some extra suggestions for that as well.
Did you try the recent 13.x release of Misskey? Apparently it got a massive speed improvement. Didn’t try it myself yet though and also don’t really have fresh memory to compare it against.
Then maybe I wasn’t very lucky with my instance choice (libranet.de), because jfc, doing anything, from sharing a post to simply browsing the timeline, was a chore
Worth noting (since this isn’t mentioned anywhere in the text): Friendica is slow as fuck
@rom
I first struggled with my single-user Friendica instance after I added all followed accounts from Mastodon. Very slow, sometimes even not responding at all.
After tweaking some settings the site is running quite smootly. So it depends on the setup.
Yes, Mastodon out-of-the-box is faster on most machines, but Friendica can be made very usable.
Via clients like WhaleBird or Fedilab, I don’t feel any difference.
@serenity
@serenity @zeitverschreib @rom What sort of tweaks did you do?
@goatsarah
Let’s see…
PHP memory limit up to 2048
Number of workers to 20
The workers now run every 5 minutes instead of 10 (cron)
Increased the values of pm.max_children , pm.start_servers, pm.min_spare_servers and
pm.max_spare_servers.
I’m running this setup on Yunohost in a VM with 2 cores and a total of 16 GB RAM. Currently only 7 or 8 GB are used, that’s including Mastodon on the same server.
Giving the friendica instance more resources really helped, the standard settings are too restrictive.
@serenity @rom
deleted by creator
@rom @serenity eh, it’s not blazing fast but it’s quite usable and definitely beats Misskey/Calckey in that department
@serenity @f00fc7c8 @rom It’s not even sorta fast, but I’ve found it to be the most functional for me out of the bunch. In addition to Friendica, I’ve played around with Diaspora, Mastadon, Pleroma, Misskey, plus the forks Calckey and Akkoma. Friendica’s ability to easily follow tags and rss feeds, plus the lemmy and diaspora integration, as well as the ease of import/export really shows where some of the others are lacking. That said, UI/UX customization in Friendica leaves quite a bit to be desired.
@serenity @f00fc7c8 @rom This is the kinda speed problems I see with Friendica that I don’t see with other AP interfaces.
6 days ago (Received 18 seconds ago)@eshep @serenity @rom That’s not always because of a speed issue though. I usually see that on messages that were boosted or searched recently, in which case it’s no different from other AP servers.
@serenity @f00fc7c8 @rom Good to know. I’ve noticed it mostly on post pulled in from hashtag follows. I thought maybe it could be something to do with how those are polled.
@eshep @serenity @f00fc7c8 @rom If you’re following a hashtag you’ll see posts from people nobody on your instance is following and those won’t come in as they’re posted, but will be fetched as a side-effect of actions by people that are being followed.
@serenity @f00fc7c8 @rom @eshep Exactly. Seems there’s still a lot of confusion about how federation works.
My issue with Friendica is that it takes several seconds to render the “timeline” sometimes. Tempted to move the docker containers from the RPi that’s currently hosting them to my M1-pro MacBook.
@serenity @rom @eshep @f00fc7c8 In my experience it is not because Friendica is slow. then Its because someone I follow just commented a post and then the server had to download the post whereas before there wasnt a reason to because i didnt follow the OP.
@serenity @rom @eshep @f00fc7c8 I tried to import/move my profile to my own instance but I got an error saying page not found :S did you succeed with that`?
I agree with what you say. Is your experience that Friendica is slower than the others? My guess would be that Friendica is programmed in PHP which executes slower than other programming languages.
@eshep @serenity @f00fc7c8 @rom @anders I am not sure it is about PHP but it depends how you see it.
Friendica can be rather resource-intensive, especially for public instance. Some of them feel slow because they are somewhat overloaded. On badly overloaded servers delivery times suffer.
Also mostly static interface of Friendica which I like also means most operations reload the entire page. E.g. when you go to someone’s profile it can take the same time (say 8 seconds) on Friendica and Misskey/Pleroma/Mastodon but on the latter you don’t see the page redrawn - it stays the same and something spins loading the data. On Friendica you see full page reload. It is perceived slower.
Also in some aspects Friendica is inherently slower by design. E.g. when posting comment on Mastodon it is fired off right away (well, almost), on Friendica it is queued for delivery and the worker runs on interval (usually every 2-5 mins). It isn’t a problem but feels more like email or old school forum while other platforms can feel more like instant messaging.
@shuro
@serenity @rom @eshep @f00fc7c8 @fediverse
Good observations. Another thing about full page reload is that it will also put more load on the server.
On my Friendica server I am running the Daemon instead of Cronjob so for me it gets sent immediately :)
@eshep @serenity @f00fc7c8 @rom @anders Interesting, I didn’t know about daemon. Can you share how you have it configured? I set my server some years ago, I think it didn’t have this option back then but I’d like to try.
@shuro
@serenity @fediverse @rom @eshep @f00fc7c8 @shuro
You start the Daemon with this command: cd /var/www/html && php bin/daemon.php start
If you don’t have your Friendica installation in /var/www/html, change it to the proper location. You can run the command at system boot.
@eshep @serenity @f00fc7c8 @rom @anders Thanks, I’ll try this and see how it compares with cron!
@serenity
@fediverse @rom @eshep @f00fc7c8 @shuro
I assume that you don’t run it through Docker since you mentioned Cronjobs. Let me know if my assumption is wrong because then I can give some extra suggestions for that as well.
Did you try the recent 13.x release of Misskey? Apparently it got a massive speed improvement. Didn’t try it myself yet though and also don’t really have fresh memory to compare it against.
Then maybe I wasn’t very lucky with my instance choice (libranet.de), because jfc, doing anything, from sharing a post to simply browsing the timeline, was a chore