• Fur Affinity Forums are governed by Fur Affinity's Rules and Policies. Links and additional information can be accessed in the Site Information Forum.

FA running slow?

Kanga274

New Member
N3X15 said:
Kanga274 said:
Another Suggestion, Asking people, To delete all Old Journals

Another suggestion, Asking you, to stop speaking, like WILLIAM SHATNER! D:


Much of the problem is that the site's database is poorly designed and stuff. It shall change soon enough.

No need to be rude
 

uncia2000

Member
Kanga274 said:
N3X15 said:
<clip>
Much of the problem is that the site's database is poorly designed and stuff. It shall change soon enough.
No need to be rude

Both parts of that first post are true enough, to be honest.

One of our major issues at present, database-side, is for spawning the notifications when community members submit their journals.
And, of course, large numbers of people with hundreds or thousands of watchers have been doing precisely that, in the past few days.
Very easy to have a cascade problem with those, then users log-jamming behind opening multiple tabs since they don't get a response from the first, etc., etc.

Early this morning 0500 EST, I was getting 0.08 second times on my page: now, it's... well, 30 seconds... more.... :?

Does need to be fixed and will be fixed.
A half-decent response time is one of the fundamental requirements for FA.
 

Aquin

Haunted
*nods* Its happening to all of us i think. Hell, i feel like im on a 56k modem when i browse the site. Ive been pretty patient about it. The real annoyance is when i try to upload a new track. Sometimes it doesnt even load at all on the first try.
 
Summercat said:
I just wait until late in the day. Then it goes away.

That all depends on where you are. By 5 am in the morning easter time it started to clear up decently but the main page STILL took ages to load. My page didn't take long, and I could get ot my messages quickly, but on the front page it was sluggish as can be. :
 

SpiritCreations

Wanted Dead or Alive
SO SLOW! It's been ticking me off too but it happens. It takes me approximately 144+ seconds to load a page. I have DSL. I think I remember FA being slow BEFORE the drama, though. Last night when they did an upgrade or maintenance or whatever.. it was really fast for awhile but then it slowed down again. Pity. :(
 
XD Now to add to the issue sorta.. it keeps telling me I have one submission to look at but I can't, for the life of me, see it. At all XD Ah well. <3
 

Aquin

Haunted
kuronekotenshi said:
Summercat said:
I just wait until late in the day. Then it goes away.

That all depends on where you are.  By 5 am in the morning easter time it started to clear up decently but the main page STILL took ages to load.  My page didn't take long, and I could get ot my messages quickly, but on the front page it was sluggish as can be. :

Easter time?! I cant wait THAT long! xd.

j/k i know what ya ment.
 

johnofe

New Member
i got fast internet and it can take many minutes to go from white screen to and FA page.
FA is just totally sluggy, the only reason i stay with this site is cos its content and people are great.
why doesnt FA use the same code (similar i mean) to fanart-central.net or deviantart.net, theire sites dont grind to a hold and their search still works!
 

SpiritCreations

Wanted Dead or Alive
One odd thing is that now FA is running a lot faster. It's NOT what time you get online. For the past... week or so it's been running slow for me 24 hours a day. Also, deviantART has been slow for me in the past as well.
 

Kanga274

New Member
Here is how to help FA run a little faster But its what you can do to help FA. Its by Deleting all read messages yu dont need anymore and all Old Journal Entries. I believe if you do this. FA might run Faster during the end of the day. Now most of you might live in another part of the world then me. I live in the US. So its durning the afternoon to evening time that FA lags badly. No one can post anything at all cant see anything at all. So, if you could help FA out, by deleting any old mail you dont need and any old journals. Thanks for reading people, and possibly Helping too.
 

nrr

Member
Warning: slightly ranty.

uncia2000 said:
Does need to be fixed and will be fixed.
A half-decent response time is one of the fundamental requirements for FA.
I kept mentally giving you guys two months to fix things. You know, as a benefit of a doubt sort of thing.

... but that always kind of falls short and leaves me disappointed and frustrated. I would kind of like to see FA suck less, but there isn't a really apparent desire to fix anything. I've seen nothing but side-stepping since December. What gives?

Sure, there's lots of talk of hiring (well, bringing on; hiring implies monetary compensation) programmers with at least a BSc in computer science and whatever else, but that really doesn't help things. Bearing a CS degree or not, the programmers need to understand that FA's codebase needs a lot of TLC and refactoring done in order to become at least the slightest bit sane, and unless your programmers know how to shift between design paradigms in various, potentially unrelated chunks of code (while also documenting what is going on -- use Subversion and something like RT or Bugzilla!), how is FA's technical operation going to become even marginally better?

Also, you need to start looking at people who've done supercomputing work with regard to Web applications and file/database servers. If FA does end up growing any more, having only one machine running things won't suffice, and you'll have to start branching out. If I didn't think it'd look tacky to have this whole paragraph italicized, I'd do it. I absolutely cannot emphasize this point enough.

That said, a lot of the slowdowns on FA are not really a result of the codebase itself (though, I guess it could be said that it's an indirect result...); rather, they're a characteristic in increasing load on MySQL and Apache/PHP both. Both pieces of software are heavily CPU-bound, and MySQL adds in the complexity of actually being I/O-bound, with most of its operations thrashing the CPU being related to fetching things from and storing things to disk, so there's competition going on as to who gets more attention on the CPU, and the kernel's scheduler is doing the best job it possibly can. This can be alleviated temporarily by shoving MySQL off to another machine and focusing on good ways to increase I/O throughput for data being fetched from the database.

Ask your programmers how to implement this too BTW. That's an exercise for you guys; it's an easy $700-800 in the pocket for me.
 

furryskibum

Member
The techie side of me really enjoyed reading that, nrr, even though I know almost nothing about programming. XD

Sounds like a practical path to a solution, though!
 

uncia2000

Member
nrr said:
I kept mentally giving you guys two months to fix things. You know, as a benefit of a doubt sort of thing.

... but that always kind of falls short and leaves me disappointed and frustrated. I would kind of like to see FA suck less, but there isn't a really apparent desire to fix anything. I've seen nothing but side-stepping since December. What gives?

I'm fairly sure you know the answer to that, nrr. Thus rhetorical?

nrr said:
Bearing a CS degree or not, the programmers need to understand that FA's codebase needs a lot of TLC and refactoring done in order to become at least the slightest bit sane, and unless your programmers know how to shift between design paradigms in various, potentially unrelated chunks of code (while also documenting what is going on -- use Subversion and something like RT or Bugzilla!), how is FA's technical operation going to become even marginally better?

Also, you need to start looking at people who've done supercomputing work with regard to Web applications and file/database servers. If FA does end up growing any more, having only one machine running things won't suffice, and you'll have to start branching out. If I didn't think it'd look tacky to have this whole paragraph italicized, I'd do it. I absolutely cannot emphasize this point enough.

That said, a lot of the slowdowns on FA are not really a result of the codebase itself (though, I guess it could be said that it's an indirect result...); rather, they're a characteristic in increasing load on MySQL and Apache/PHP both. Both pieces of software are heavily CPU-bound, and MySQL adds in the complexity of actually being I/O-bound, with most of its operations thrashing the CPU being related to fetching things from and storing things to disk, so there's competition going on as to who gets more attention on the CPU, and the kernel's scheduler is doing the best job it possibly can. This can be alleviated temporarily by shoving MySQL off to another machine and focusing on good ways to increase I/O throughput for data being fetched from the database.

Ask your programmers how to implement this too BTW.

Thanks for the freebie feedback/input, nrr. Appreciated.
 

nrr

Member
furryskibum said:
Sounds like a practical path to a solution, though!
It's what happens everywhere with larger Java Web applications, and it's what happened with LiveJournal as they started growing too large.

The difference with the enterprise Java stuff compared to comparable open source offerings is that the tools for clustering are built into the platform as a neat little prepackaged framework, and the big commercial relational database packages (and their respective connectors in Java) all support managed clustering internally. This makes scaling up a particular application in most cases as painless as plopping a new configured node in the rack.

In the open source world, this stuff is only just beginning to see any sort of coherent adoption. Brad Fitzpatrick wrote DBI::Role for Perl to help LiveJournal adopt the concept, but he hasn't really released it to CPAN yet, and MySQL has only just begun (compared to the commercial players) to get the replication thing down in order to make clustering with it somewhat feasible. Regardless, it's been in production use over at LJ for three years, so it's not really anything that can't be done. It's only painful to get it right.

Edit: If you'd like to poke around at some of my work-in-progress framework code for PHP: https://code.corvidae.org/svn/NRRWebFramework/trunk/

I have some classes prefixed with NRRMySQL that may be particularly interesting. NRRMySQLRole isn't even close to being done, nor has it actually been shoved through PHP to make sure things work.
 
Top