Redundancy: an unnecessary luxury?

Do you ever stop to think about what can go wrong if you don’t receive any email for a few hours? Of course, plenty of emails can wait a while. But if a customer reports a fault, you want to fix it as quickly as possible. If a big enquiry comes in, you would rather call back immediately. And if you receive an interesting offer by email, you need to be quick off the mark.
A technical fault can then really throw a spanner in the works. By the time the engineers have arrived and the problem has been fixed, six hours may easily have gone by. With perhaps more than one missed deal and angry customers as a result.
Continuity guaranteed
Unfortunately, things do break sometimes. That’s why Onlinespamfilter has built redundancy, or spare capacity, into every possible bottleneck in the process. As soon as a system fails, we automatically switch over to another system, so your emails still always arrive within our target time of 10 seconds at most.
For example, our servers are spread across two data centres in different locations, linked to each other by a dedicated connection. Each data centre is connected to a different internet provider, and we buy our domain names from different DNS registrars. Behind the firewalls, several spam and virus scanners are at work in each data centre. All data is stored in real time in a distributed database at different geographical locations. And we do all this at low cost, because we work exclusively with open-source software that we adapt ourselves where necessary.
What’s more, our servers run at no more than 50% of their capacity, so an entire data centre can drop out without any noticeable effect on email handling. An added advantage of a system that simply keeps running is that we have the peace and quiet to solve the problem in a controlled and well-considered way.
How safe is safe enough?
We look carefully at how many backup facilities are needed. Two internet connections are enough for us, for instance, because it almost never happens that both give up at the same time (it has certainly never happened to us!). But for database nodes, three is the minimum. Otherwise you don’t know which of the two has the correct data if one of them has been temporarily down.
These double and sometimes even triple measures do make the system somewhat more expensive to maintain, but when we handle the email for well over 2,500 domains, we would rather not take any risks!
Pssst.. What about your own redundancy?
If this blog has given you the idea of bringing in a second internet provider yourself, it’s good to know that we can deliver your emails over multiple connections. If one provider has an outage, you automatically receive them via the other.
