The attempt at the Guinness World Records® Record for most comments on an online news post in 24 hours is set to start at 6am on Thursday 5th January. This is going to be an interesting experiment which will hopefully raise a lot of cash for the RNLI.
In order to make a success of the record attempt we have moved the blog off its old shared platform onto a new state of the art job hosted at Timico’s new data centre in Newark. This is before the data centre is officially up and running so we have all its resources to ourselves. I am heavily indebted to Timico for this.
We have been quite busy getting the new infrastructure up and running so I am also very indebted to neighbour Steve Nice and his company Forlinux for their help with the platform software.
The basic architecture looks like the diag on the right (click to enlarge). A load balancer fronts a number of wordpress servers that in turn feed off a two database (one replicates to the other as a backup).
Each server is a Cisco UCS B200 blade and connects with 2 x 10TByte EMC Storage Area Network (SAN). The blades are dual 6 core (hyper-threaded) Xeon processors with 96Gigs of RAM each and with the Hard Drives disabled for mtbf optimisation purposes. The components are all connected using a 10GigE backplane.
The data centre initially has 2 x 10Gig connections coming into it so there is plenty of bandwidth available. The site supports IPv6.
We are using WordPress 3.3 on top of CentOS 6/Apache 2.2/PHP5.3 and mysql 5.1 – version numbers are hidden from people querying the web server (Apache).
The blog itself, already relatively light on plug-ins, is based on the Thesis theme and has been pared down to the minimum to maximise performance.
Some plug-ins are worth a mention. Caching is employed using WP Total Cache for performance optimisation. Simple Twitter Connect is used particularly for comment tweets and AddToAny is used in the post for sharing.
We were kindly offered CDN capacity by Limelight Networks but we as we have already been working overtime to get the data centre ready there wasn’t time to incorporate this. The CDN is probably just a “nice little extra” on this occasion anyway as the content isn’t particularly heavy on bandwidth.
If you want to help with this world record please tell your friends – the post is going live at 6am on Thursday at http://www.trefor.net. If you regularly use a forum or have your own blog please leave a post with a link to this site. If you want to donate please use the JustGiving page or click on the big green “DONATE NOW” button in the sidebar.
13 replies on “The tech detail of the platform being used to host the world record attempt #comment24”
FYI For me your header image is ‘timing out’ from the ‘staging.trefor’ addy – hopefully someone’s just tuning things 🙂
Thanks HmmmUK – we moved the platforms across this pm. This may be DNS cache related but will ask someone to check it out. Thanks again.
I would suggest not using Apache! It is very bad for lots of connections unless heavily cached outside using something like Varnish.
Switch the web servers across to lighttpd or nginx. We had to do this for http://planefinder.net and CPU/memory (and servers not falling over) went down drastically!
Thanks Lee. Appreciate the input. We did discuss not using Apache but the general consensus was to keep it. I will however revisit it with the guys in the morning.
All the best
Tref
I think you may run into issues as you need pretty much live uncached (as much as possible) due to the comment threads.
We went from being floored with 500 concurrent users per server to them now being able to cope with 1 server if needed!
hmmmmm anyone running a book on if the site will crash? Lee seems quite concerned
@ rupertrogue
If I recall correctly, it’s impossible to give a definitive answer on how many connections Apache can cope with – as it all depends on the type of load. A static page will consume less resources than a PHP page (which is heavier on resources). It also depends on how you’ve set things like PHP up…
nginx and lighttpd are much better than Apache at handling lots of small requests, and serving up static files (I think due to clever caching stuff…). A few of my friends have put nginx in front of Apache to serve their static files – leaving Apache to serve the processed files… Some of my friends also use varnish to cache stuff.
So I guess in this case – nginx would serve up the static page, and when you add a comment, it gets posted back, and passed to apache? At least that’s my understanding…. I don’t really get servers.. javascript’s my thing 🙂
In any case, the ForLinux chaps have set the software up – and they know their onions. If they say Apache, then they’ll have a really, really good reason for it 🙂
I can’t wait to see what happens 😀
Will you be publishing any traffic/load graphs? cacti/mrtg etc
Nat – yes definitely will be publishing that kind of data.
Tref
I’ve just mentioned this article on Twitter.
Any visitors – make sure you leave a comment on the World Record attempt page for the RNLI!
http://www.trefor.net/2012/01/05/world-record-attempt-for-most-comments-on-an-online-news-item-in-24-hours-has-started-comment24-rnli/
A good way of testing that those blades are sharp enough.
Good to see those Cisco Servers being put to good use !
Would love to see how much bandwidth got used today.
Seems ashame there weren’t people sitting in the comments threads having long conversations!! That would have racked up nicely.