Discussion in 'Server Configuration and Hosting' started by billybatz9, Nov 8, 2014.
I am just curious as to who uses hostgator?
We moved from Bluehost to HostGator, from HostGator to XF Host, and from XF Host to @Mike Edge.
From the little time we were on HostGator, it was awful. The site was hardly ever up.
So basically you moved from EIG to EIG. Same owners Horrible company.
Bluehost wasn't as bad, though, until a little earlier this year (even then they weren't that bad), and we had them since the site opened in August 2011.
Nobody who wants a well-perfoming, cost efficient site.
Who do you use?
I have a server with QuadraNet that I have installed ProxMox on that allows me to create VPS's.. in addition I created a VPS (16G RAM, 900GB disk) that I have cPanel installed on that allows me to create accounts like shared hosting providers do.
I know you didn't ask me. But me personally I run a VPS off Digital Ocean for $10 a month. Running NGINX as my web server. I have it configured well enough that with OpCache enabled I can throw a 250 user /1000 concurrent connection load from loadimpact.com at it and it doesn't even blink.
I actually have memory usage down far enough that the $5 a month plan would likely handle the same load test which is much more traffic then even my gaming communities site gets normally. So with decent tuning, a little cheating, you can have a custom VPS for shared hosting kind of money.
I'm envious of that, my user's content ends up being ridiculously bbcode intensive and long which really doesn't help performance.
Memory isn't so much of an issue, as having enough CPU cycles to throw at it.
Thats where opcache made the difference. I can't say so much towards the load bbcode parsing will cause but this was what happened for me.
Out of the box Percona 5.6 was using 450 mb and with my config it was using 700. Memory ran out very quickly. I have the initial allocation of MySQL down to 75mb of ram and my innodb buffer set around 1/3 of a gig. It expands way beyond 75mb because of that of course but really it never goes past 400mb of ram right now. The reason I cared was php-fpm workers were eating up ram causing mysql to swap.
With that solved I noticed with 1 thread I could do 50-75 VU's before CPU load reached 100% and then with 2 threads I could serve maybe 150 VU's in a test before 100% load. Tuning down the number of workers a little and OpCache enabled greatly reduced the CPU load and now the load test can hit 250 VU's and 1K connections just fine and its snappy on my end during the test.
I tested a backup of TBG's website on it and I am getting more consistent times than I do with the shared host. But the performance figures didn't change much. The shared host is a virtualized instances with 3GB of ram though so it has more to work with. Not bad for an idiot I suppose.
While the average post is ~120 characters, having commonly frequented posts which have hundreds of links and thousands of words with heavy formatting is fairly common on the first page of a number of threads. Worse, I think I've got some addons which are hooking the "load_class" event without any type hinting.
It is quite a different workload from most other bits.
I could have sworn I saw an option in the acp for caching bbcode. Have you tried using that?
I am using it, and it really helps. It is why some pages render in ~1-2 seconds rather than ~8-12 seconds.
Without caching of the final output, bbcode -> html conversion is quite expensive for long posts with complex bbcode. Especially as I have a few addons which implement some event hooks without using hints.
8 -12 seconds? maybe you should issue some guidelines on acceptable amounts of formatting
Separate names with a comma.