By Griffin on Friday, 22 August 2014
Posted in Technical Issues
Replies 4
Likes 0
Views 834
Votes 0
We have been building a platform on EasySocial for about 6 months and have done some Alpha testing with 200 users recently. (Not concurrent users, just 200 users from around the world using the platform probably with a maximum concurrency of about 20.)

My questions are:

1) How many users have you successfully hosted on your EasySocial instance?
2) How many concurrent?
3) What's your secret? :-) Or Secrets?! (CDN (which one?) Hosting architecture? Load Balancing? Avoiding certain Joomla! process hogs? Porting to MongoDB? You name it!?)

A few years back I managed to support about 300 concurrent users at decent performance levels on JomSocial running on two maxed out servers (one for DB and one for Process) using CDNs (content distribution networks) to handle heavy files.

Overall I was disappointed with JomSocial's ability to scale and I have big hopes for EasySocial.

I'd love to have some anecdotal evidence so that I can move being "hope" and make a real business case to my co-workers for the platform.

Cheers!
Grif
I've had thousands of concurrent users on Jomsocial on a single server, no CDN, with no problems. I been a Stackideas customer for over a year and I have tested every version and had some other testers but I have yet to implement EasySocial on a production site. I have been waiting for 1.3 to go with a production site. That said I expect it to be superior in every way to Jomsocial...

My opinion on CDN is that it is useful for 1) ubiquitous elements like jquery and bootstrap. 2) sites with a diverse geographic audience. However I have witnessed CDN actually slowing sites down. It is not a cure-all nor is having a blazing fast multi-core server with tons of RAM. I've had slow Joomla sites that were on dedicated servers on tier 1 networks and the server(s) resources were barely being used - but site still slow. How you optimize is what makes the biggest difference. Find your bottlenecks. Often with Joomla it is either mysql or how php is playing with mysql. Poorly written scripts, databases lacking the right types of indexing. Minimize what you install in the way of extensions. Joomla has always had a ridiculous amount of calls. The number of calls is actually a big factor. Everything you install is adding to that. So you have to get the other stuff right. Image optimization, caching methods. Optimal web server (apache / nginx) settings that work best for CMS. php settings. innodb optimizations. When 1.3 comes out my first test will be on dedicated server with nginx and mariadb so in this case it will be xtradb rather than innodb. we'll see how it goes.
·
Friday, 22 August 2014 11:25
·
0 Likes
·
0 Votes
·
0 Comments
·
Hey guys,

To be honest with you, we have never profiled any sites with such huge concurrent users before but we would definitely love to help profile this if you have a large site with huge concurrent users. We could work out a deal if you guys are interested. I am looking forward of having the following:

1. Admin access
2. Server access
3. FTP access
4. Total number of concurrent users
5. PHPMyAdmin
6. Let us know when is the peak hour (with timezone too)

If you have a site that have more than 250 concurrent users during peak hours, please get in touch with us at https://crm.stackideas.com and I personally would like to have a chat with you guys to see if we can use the site as a benchmark
·
Friday, 22 August 2014 14:04
·
0 Likes
·
0 Votes
·
0 Comments
·
Hi Paul (& hi Mark),

I wondered if Paul had taken the production plunge with 1.3.12 and if Mark had profiled any sites with large user bases. Mark, if you're interested we'll be running another test with a few hundred testers in Feb and March.

Paul's comments on utilizing the XtraDB engine are interesting. Anybody else been successful in scaling performance this way?

Grif
·
Monday, 12 January 2015 03:19
·
0 Likes
·
0 Votes
·
0 Comments
·
Hello Griffin,

Sure, we'd be happy to run some performance test on your site
·
Saturday, 17 January 2015 15:49
·
0 Likes
·
0 Votes
·
0 Comments
·
View Full Post