Smooth performance through multi-server hosting

Last week, we successfully passed the first high stress test with one of our clients on Codesphere. After some preparation with stress test scenarios and some smaller events we can proudly say that even larger high traffic events with resource intensive streaming and live chat content is not a problem for our infrastructure.

Let’s take a short step back and look at the backgrounds. You will see why we consider this a major achievement.

1. We have larger sites with generally relatively high traffic hosted on Codesphere. These are mostly larger e-commerce sites built with WooCommerce.
2. We have seen our platform handle traffic peaks during Black Friday just fine
3. Our whole goal is to make a truly scalable multi-server infrastructure easy enough to access and setup without needing DevOps engineers
4. So far we had not gotten the opportunity to prove just how scalable it really is though

Well until last week anyways. Our client Female Leadership Academy provides educational services for female business leaders and those striving towards leadership positions. As part of this they offer courses both offline and online. Apart from these closed audience courses they also host online events that are open to the public and attract larger crowds of people. We are talking about live streaming and live chat interactions with thousands of people online at the same time resulting in more than 100 requests per second - quite the heavy load for any hosting.

Their website(s) are built with WordPress and with the previously available hosting (Raidboxes in this example) they always had to cap the number of people that can register for these events, because the hosting simply would collapse at a certain point with too many people online at the same time. With Codesphere they were able to set up a scalable, multi-server infrastructure with ease, maintaining great page speed even under load.

Not only can they now handle much more visitors for these events, they also have full flexibility over their hosting capacity. Since they know the time of the events in advance they can provision ample computing resources for the event and after the event they can simply reduce the amount of servers back to normal load which helps keep their cloud spend to a minimum.

Let’s take a closer look at the numbers of this specific event:

• 3000 registered participants
• >120 requests per seconds during the peak

Translating this into Wordpress (PhP) terms, each request occupies one kernel and takes around a second to process. It’s rather simple math to see how balancing the load between multiple servers will help to keep a smooth performance for users. 100 requests per second with computing time 1 second each would mean a single server with 4 cores would already result in load times of up to 25 seconds. For this event we were able to keep this within acceptable ranges (<10 seconds) even during the absolute peak. Also adding additional servers for even more requests per seconds can be done with a few clicks.

The setup chosen was able to handle the requests smoothly and the event was a success. For us this is really reassuring as our flexible setup allows hosting these types of events with reasonable computing costs. For our next release we are pushing an update that will automate this process (autoscaling replicas that take less than 5 seconds to launch) which will make this easier even for less foreseeable traffic peaks 🚀

Want to learn more? Talk to one of our solution engineers now!

Stay tuned for more exciting updates coming very soon and check out our website if you haven't done so yet.