Before we take over, the visitors of the site were experiencing some major problems related to the server setup. Now, those are long forgotten. As a result, we have achieved zero downtime, meaning the site is without any service interruptions. Additionally, we have configured a load balanced setup in order to resist ddos attacks and handle traffic peaks without hiccups.
Choosing the server provider was an essential part of the process. After careful considerations, we decided to use “OVH” as the main server provider and “Docker” as the main deployment strategy.The original server provider for the Fuller Treacy Money’ site was “Softlayer”. Switching to “OVH” reduced their overall costs by 60%. Needless to explain further, the numbers speak for themselves.Moreover, the new setup with Docker made the setup highly available as well as load balanced.
The next change that followed in improving the performance of the site was to switch from apache to nginx with the page speed module. Then, we changed the MariaDB for the db. With some additional tweaking, the page load was reduced by an average of 1sec (from 3 to 2 sec) and the maximum throughput was doubled with the score of 60 pages per second measured with ab (apache benchmark). The overall setup had many moving parts so dockerizing everything was challenging but not impossible. The final setup is a system that is very reliable and very easy to deploy and upgrade.
We chose to use 3 servers on “OVH” spread across Europe and Canada. Why would we do that you might ask, well the answer is pretty simple – the servers are region isolated. In case the entire data center in Europe crashes (nature disaster), the data centre in Canada will not be affected. We ensured that the whole system is backed up and in case of unforeseen disaster (although we can’t imagine such), which takes all servers down, the new server can be deployed from the backup system in no longer than an hour.
Using the “OVH” system is a first for us. Some online reviews are not as positive as one would imagine, but after multiple tests, we know for a fact that our system has no single point of failure. Therefore, we fully trust our choice.
The 3 servers are connected together with the new VXLAN overlay networking introduced just recently by docker. The data and the database are fully synchronised between the servers so they are identical at all times.
Since logging is really important against debugging and troubleshooting, all servers send the logs to third part logging collector. For now, we decided on https://sematext.com/ for the logging collection, but it might deploy our own solution in the near future to bring the cost down in the long run.
The monitoring as well as notifications are set on http://www.site24x7.com/ which connects to the cloudflare’s api. When the server is down, it automatically disables the dns record to the sick server and it restores it when the server is up again. We find that to be a very low cost solution without using a front load balancer that guarantees no visitors will be send to the server, which is marked as down.
Final words for this particular project will be that we gave our best to optimize the performance of the site and ensure that we will achieve near 100% uptime. We are proud to work with the Fuller Treacy Money company and we are grateful for trusting us with their business. Vip Consult strives as usual to provide the best quality service. We enjoy working together with companies from different sectors and help them grow.
- Automated deployment in under 1h
- Reduces running cost by 60%
- HA and LB - expecting near 100% uptimes
- Increased maximum page handling from 15 to 60 pages per second
- Decreased page load times from 3 sec to 1 sec