Assuming you are serving static content, a much easier method is to use a CDN for all your assets (including the WP generated pages). It might cost a little more for big spikes, but it makes you completely immune to almost any amount of traffic for very little effort.
10 Million hits a day with Wordpress using a $15 server
91–100 of 115 posts
Re: 10 Million hits a day with Wordpress using a $15 server
#92Earlier quoted context omitted.
I think "a $4217/mo CDN bill" may be a bit of an exaggeration :p I honestly had no idea how much CDN's charge, threw some rough stats into AWS Monthly calculator for CloudFront (596GB/Month, 500kb avg. object size) and the bill was only $73.87 /month. Call me surprised!
Sounds about right to me. How did you get to 596GB? That seems very low for 300,000,000 visits (remember this was a daily figure, but you're paying by the month). If you have 10 requests per visit (reasonable for the low end) that turns into a cool 3B requests. That itself is $2,250 a month. A million visits times 500KB = 476.84GB. Per day, so 14TB per month. At best that's another $1,600 a month.
Epic math failure on my behalf. Calculated based upon 10M visits per MONTH! D'oh!
Re: 10 Million hits a day with Wordpress using a $15 server
#93FWIW: Amazon's Micro instances will burst to saturate unused CPU cycles on the host machine up to a hard-capped limit before being choked to death by the hypervisor. Had he run the Blitz test for 10 mins or more you would see the spike in traffic up, beyond what you think a Micro should sustain, and then it plummet to near 0 for a disturbingly long period of time[1]. If you are unlucky enough to have a Micro on a hos…
After reading this my question is: will his system work on a small instance on ec2 instead of micro? It sounds as though the majority of issues you're describing are micro-specific and while there's a thousand "tune wordpress" blog posts out there it sounds like this one is a good idea minus the use of a micro instance.
Re: 10 Million hits a day with Wordpress using a $15 server
#94We're including a tool in Ubuntu called juju [1] that will enable us to ship cool setups like this to users.
We did something similar for Wordpress [2] and plan to ship it in 12.04.
I'd love to bring you on board so we can compare configs and make it even better. Also rkalla is correct, micros are great for prototyping but you'll need at least smalls for production. The nice thing about micros is that you can set it up, tweak, and then later reboot them into larger instances, so they're nice for playing, but I wouldn't run a live site on them.
[2]: http://www.jorgecastro.org/2012/03/18/redeploying-omg-ubuntu...
Re: 10 Million hits a day with Wordpress using a $15 server
#95Or leverage a template cache on (insert any framework here) and get 2,000+ hits a second on a Linode, or 172,800,000 hits day.
Re: 10 Million hits a day with Wordpress using a $15 server
#96tl;dr - Varnish The only way to make WP "fast" is to make it nearly completely static. Get a half dozen people working admin area and your server will cry.
Though people on WP doing admin, if you have 6 people, are unlikely to be doing things every second - most of it is writing stuff, looking at things, so even if they were doing a click every second that's still 6 requests per second that for even completely non-cached dynamic stuff is completely irrelevant.
Re: 10 Million hits a day with Wordpress using a $15 server
#97Or leverage a template cache on (insert any framework here) and get 2,000+ hits a second on a Linode, or 172,800,000 hits day.
Probably closer to 20,000 hits per second with Nginx configured correctly and if the framework doesn't impose too much overhead. Alternatively, use ESI with Varnish and the HTTP Cache spec and get 10,000+ hits per second on a relatively dynamic website (read: micro-caching).
Re: 10 Million hits a day with Wordpress using a $15 server
#98Earlier quoted context omitted.
Your forgetting that this favors your more than them. People dont want to learn a new system, and they are confortable with wordpress. There are millions of devs and designers also familiar with wordpress - which can mean they can leave you in an instant and get anyone else to make changes or add feature to the site easily.
That's assuming they are already familiar with wordpress, most people have never used a CMS in their life. If requirements are simple then all you generally need is an admin area with about 3-4 links. Something like: Change Homepage Text Show Customer Inquiries Add Item to catalog This is much easier to understand to a newbie than "Add Page" , "Add menu item" etc. If they want complicated changes they will usually en…
* Find a new developer to work on my site
* Install a plugin to handle facebook like buttons
* buy a new template/design for $40
The first one is key though -- as a business decision.
Re: 10 Million hits a day with Wordpress using a $15 server
#99FWIW: Amazon's Micro instances will burst to saturate unused CPU cycles on the host machine up to a hard-capped limit before being choked to death by the hypervisor. Had he run the Blitz test for 10 mins or more you would see the spike in traffic up, beyond what you think a Micro should sustain, and then it plummet to near 0 for a disturbingly long period of time[1]. If you are unlucky enough to have a Micro on a hos…
After reading this my question is: will his system work on a small instance on ec2 instead of micro? It sounds as though the majority of issues you're describing are micro-specific and while there's a thousand "tune wordpress" blog posts out there it sounds like this one is a good idea minus the use of a micro instance.
The techniques Ewan outlines are spot-on; my clarification is for people to have the right expectations of the Micro's, not on the information he provides on tuning a WordPress install.
That said, Smalls and up on EC2 will give you more consistent performance (they aren't built to provide the spike-performance Micros are) BUT are still relatively subject to "noisy neighbors".
The smaller the instance, the bigger the impact felt; it is still possible to have a small grind to a halt because of a noisy neighbor and typically I do not see people hosting web front ends with smalls because of the erratic and poor performance (in general). Possibly a fleet of them behind an ELB, but even then I see Mediums and Larges much more often as the "web server".
The mediums provide an awesome (and typically unsung) balance between monthly cost and performance on your way to a large (which can get quite expensive).
Re: 10 Million hits a day with Wordpress using a $15 server
#100Earlier quoted context omitted.
The thing is this - configuring a server is one set of problems, building a decent CMS database/backend is another set of problems, and building a decent browser-based UX for content authoring is a third set. Very few people have all three of these skills, and it's fair to say that 2 and 3 are not yet solved problems. Rolling your own CMS for any non-trivial purpose is always something that sounds like a good idea un…
> if it's powerful enough to flex to non-trivial needs (Modern WP, Drupal, Django, CQ, etc), then it's probably going to feel like a bloated/complex mess to a programmer, because of all the nuances in the problem space. This is a good rule of thumb, but I have found one shining exception, and it is called ProcessWire. If you've never tried it, I highly recommend a look. It's a CMS that essentially offers you a blank…
+1 for ProcessWire. I've been playing with it for a day now and I love its flexibility.