Even before we got that right, it was clear to us that we needed to virtualize to be more flexible with our hosting. We just wanted to be responsible for one physical server and virtualize all projects. So we got the biggest dual-Xeon server, which we were able to raise at short notice and based on the Xen virtualization system set up a host system. We were so excited about our virtualization idea that we virtualized a good deal. In hindsight, probably a bit too much.
With the switch to virtualization and a rack in the professional data center (RZ), we had already taken an important step. Reliability and connection, and the price / performance ratio were just great. In addition, we were in good neighborhood (yes, there are still RZs in which the servers are labeled with real names) and the data center was also only 3 km away from our former office, so that one was there in an emergency also fast.
So we ran this wonderful virtual server world and were really happy with it. Until the day came when we asked ourselves how we actually update the host system, that is the Xen server, during operation, without running the risk of everything being down. Anyone who has ever operated a web server, knows that it is not a good idea to perform more extensive upgrades in live operation and the strict virtualization we were forced to turn off all virtual machines before the slightest change to the Xen server and then could just pray that they boot up again or the Xen server was still functional after an upgrade.
In short, this hosting solution was doomed to fail in the long run. In addition, our self-installed Xen server on Debian caused more and more strange kernel problems (see also my blogpost from mid-2009 ) and we did not have an adequate test system to test changes safely. It would have been possible to buy a second identical server and place it next to the live server. Unfortunately, we did not pay attention when buying the server that the built-in motherboard can only be upgraded to 8 GB of memory and we were not able to get there with our virtualization ideas and new projects plus steadily growing traffic.
Get professionals on board?
So we realized that we needed a solution to manage our Xen server. Of the technology we were still fully convinced and although we wanted to do a lot differently with our experience, we simply needed a stable Xen host server as a base. So we were searching again. The Billighoster from then had new servers in the program (still desktop hardware), but the cable fire of that time was still firmly in our heads. At other hosters there were even offers managed virtualization solutions. After closer inspection, however, there were always some hooks on all finished packages. After weeks of research and a few tests, we were suddenly looking for an individual host, a specialist for our requirements.
From the rocket researchers among the hosters
About the Sun Startup Essentials program we became aware of unbelievable machine(short * um) – a small hosting company from Berlin. At the beginning I did not really understand what they were doing and somehow an individual solution seemed to me to be too incalculable. After the first conversations with the um um guys, however, it soon became apparent that we could get a partner with unbelievable machine on board who would completely relieve us of the entire management of the hardware solution and also the management of our Xen server ,
Now we are a little over a year there customer. I like to call you the “rocket researchers among the hosters”, because they are simply extremely demanding as hosting service providers from the hardware via the data center to the connection, and a team of experts has managed to find state-of-the-art The Art Hosting technology to break down on practical solutions so that the whole thing is still affordable for small customers like us in the end.
For example, we have a Sun blade in one of * * powered blade centers with enough power through dual Intel Nehalem processor, SAS RAID, and 24GB of memory to run a Citrix Xen server for us. The data center in Berlin, where * operates its hosting solutions, is the most professional data center I’ve ever entered, and everything is designed to be extremely redundant so I think it’s over-sized for our needs – but can harm the extra Professionalism is not. At least as long as the price / performance ratio is right. We have completely outsourced the Xen server host system including backups to * um. We only care about our virtual machines. As further packages, we currently use the shared firewall solution and the 24/7 restart / reset service.
Especially the restart / reset service has already saved me a lot of trouble. At a certain point, it hurts really bad, if you are suddenly responsible for everything with a large grown website offline and stupidly as a technical contact person.
In the long run, it is simply an illusion to be fully available as an individual and to solve any server problem in a timely manner in an emergency. This year I had the situation that I was with my at this time 15 months old son at the supermarket cashier, should pack the weekly shopping / pay, he was already restless, because he finally wanted to have something to eat on the table and in this wonderful Moment of expectations reached me this worst of all monitoring SMS: “t3n.de not reachable”. Of course, I could have screamed loudly, grabbing my son and running from the grocery store to the nearest internet connection, but instead, I just did nothing like that and trusted my hosting solution.
Of course, I still look at today still promptly where the error was, but if I can not respond in an emergency, the * um-standby service tries to fix the error independently. 24 hours a day, 7 days a week. At the beginning, we worked together to compile a documentation of our hosting architecture, so that the most basic sources of error can be independently remedied by every * um employee. For this we pay a monthly flat rate and the cost for each emergency operation – also depending on the day and time. So far, an absolute fair solution, which I would like to miss no more. For me, that’s an absolute plus in quality of life.
Incidentally, we meet at least once a year to discuss our hosting development, issues, and ideas for improvement. As traffic grows and new projects come along, this is an incredibly important date for me, which of course is complemented by numerous discussions throughout the year.