Below you will find an overview of all changes made to the DualDev infrastructure since founding the company back in 2010. We hope this document clears up many of the technical and executive choices we've had to make in the past. News is noted from newest to oldest date.
November - continuing development The move to AWS has been completed to a phase where we don't have to keep monitoring the platform 24/7, and where we can finally start trusting the platform again. We still have to move several services (including loadbalancers, DNS and other things), but we've continued development of a renewed website, face and customer panel. The service we offer, will be again fast, stable and futuristic, as you're used to from us, soon. Before we continue on building the new website, we still have two of the old servers in service for loadbalancing, backups and e-mail services. These will be retired the 28th of november (at night). The loadbalancing and backups have been moved. E-mail will also have to be moved to different servers to handle the load better.
September - switch to Amazon AWS When a hacker group found our fileserver IP, and DigitalOcean was majorly lacking in giving support, we were forced to make an immediate move to the Amazon AWS environment. If we hadn't made this move, that day, we would've been consecutively down for multiple hours every day, or banned off of the DigitalOcean platform with no data at all. This force move left us with little time to manage anything else but making sure we moved all the data correctly, including extra services people have entrusted on our hands. It was a hectic time, with lots of problems annoying our customers and straining the nerves of our employees.
July to August - testing DualDev platform on Amazon AWS Having huge trouble with DigitalOcean, with servers being blackholed for three hours when getting a small DDoS, we started to look for a better solution. We quickly realized a setup on different providers, and choose Amazon AWS as the champion, electing to switch our entire platform to them when we could plan it correctly.
May - loadbalancer trouble In an effort to counter the ongoing blackholing of our loadbalancers we decided to place an extra VPS outside of the DigitalOcean network (we entrusted Vultr with this task) and added another loadbalancer in a different AMS2 network. This meant we had to inform our customers of new loadbalancer IPs and had to update all DNS entries.
January to June - starting on rebrand We wanted to give DualDev a new face, and a couple of new services for a longer time now. We finally put a designer to work and started to give the new services a definition. It's a lot of work, but we're hopeful in completing this Q1 2015. Our concern about DigitalOcean's practices of blackholing VPS's was raising by the week, with more of our different loadbalancers being disabled every day.
November - first VPS blackholed We read some reports from other DigitalOcean customers, reporting data loss and more when they received a (simple) DDoS. Concerned, we encountered the same situation, however the anti-DDoS DigitalOcean has in place simply makes your VPS completely unresponsive and unavailable to the network- in or out, for at least three (!!!) hours. Long as we didn't get DDoS'ed to much, the platform was actually doing quite well and was considered stable.
July - switch to DigitalOcean By now, we have the previous failed move fresh in our memory, but we had to try again. This time we choose to set up a different infrastructure with an extremely powerful node powering the backend and fileserver, backups going to our own hardware and our hardware still managing e-mail.
May to June - switch to CloudVPS We already had some experience working with CloudVPS by now, we wanted to get managing the hardware off of our hands and leave that to people more professionally compliant with servers, next to being able to scale better. The switch was not without its hurdles and eventually we ran into a couple of major issues we were unable to circumvent. Multiple times we moved our environment, and were then almost immediately forced to move back because of networking issues, incompatibilities with CloudVPS custom patches and a lot of more trouble we did not encounter in our (smaller) tests. Much to the annoyance of our customers, and resulting in more downtime than we wanted, we eventually dropped our efforts trying to move to this provider (who also lacked in giving any support going any further than the most basic linux support).
October to April 2013 - Testing CloudVPS Our hardware is by now having a lot of trouble processing all the traffic directed at them, so we started testing multiple VPS providers trying to see if we can relieve some of the load by having loadbalancers on different networks. However, the jump between networks caused significant longer loading times, encouraging us to choose a provider and move our entire network there.
January to December - Tweaking The platform was quite stable, we just performed some tweaks and had a few close calls concerning failing hardware parts. We noticed an increasing growth in traffic and noted this is a concern for next year. We increased memory in all powerful nodes and retired the smaller nodes completely, using loadbalancers from external providers.
December - Added additional servers The original servers we had, already had trouble processing all the traffic we were receiving. We added the two extra, extremely powerful, servers to relieve the smaller nodes. The smaller nodes now served as web loadbalancers, while the two powerful nodes served as database, fileservers and php backend. We added extra memory to one of the powerful nodes, allowing it to also handle e-mail.
April - Original Launch Date
The entire DualDev platform is running on two servers, custom hardware, colocated in Leaseweb on the high tier network and managed on premise. We're still highly tweaking the site, the platform and hosting in general.