Customers who are unable to access their data and applications due to a network failure, and their own infrastructure that is down: this is the nightmare of many hosters, cloud service providers and other digital service providers. Redundancy is therefore high on the agenda of all major users of data center connectivity. It is too risky to rely on a single connection, both within one’s own infrastructure and for service provision to customers.
However, it is not always clear how suppliers offer redundancy. How redundant is redundant? When it comes to connectivity, the connections are duplicated from location to location, all the way to the data center. But it is important to know which elements in the network are also redundant. It is also important to ensure that no infrastructure is shared at the back-end. This is certainly not a common practice yet. As far as we’re concerned, it’s time to make redundant connectivity smarter and take it to the next level.
Preparing for the unforeseen
Most (in name) redundant connections consist of two separate optical fibers or channels that both go from A to B. This is often purchased in the form of a bare fiber or wavelength. Purchasing the connection at these levels raises the expectation that the highest level of control will also be available. In practice, however, this is not always the case and it is questionable whether this is the smartest way to create redundancy.
To understand this, we must first zoom in on the risks of redundancy. Especially when hiring wavelengths, we see an increased risk of ‘paper redundancy’. People often hire from two different providers and these can be serviced or have a downtime at the same time, sometimes even due to the same cause. In theory, this seems unlikely, but there are clear reasons why this happens regularly.
There are many different providers of wavelength services, but they all use a limited number of fiber networks (renting the same routes). As a result, different providers often have to deal with the maintenance of the same network. In addition, there are disruptions to different networks because fibers often lie in the same slot or duct, or follow the same route. The question is whether redundancy by two providers will lead to higher reliability. Isn’t one provider better? 1 provider can in any case guarantee that 2 redundant routes are not in maintenance at the same time.
In addition, there are risks in the data centers, for example in the network equipment or the patch, by your own actions or by other people. Here, too, it is important to know how redundancy is dealt with in the data center. And is it possible to have multiple data centers provide redundancy to each other?
Finally, (planned) maintenance may be behind schedule, which means that unforeseen downtime has to be taken into account. We can pretend that this hardly ever happens, but in practice we know that unintentional errors are also made and that network equipment sometimes fails.
Remove the single points of failure
In short: as long as there are single points of failure in the network architecture, downtime due to failures or outages is a real possibility. This can only be overcome by installing the right fallback capabilities in the network and having control over the physical infrastructure and equipment.
True redundant connectivity is therefore fully end-to-end by default. Physical connections are geographically separated and there is full insight into which routes are used, when they are affected by a malfunction and when they go into maintenance. Latencies can only deviate from each other to a limited extent. All equipment is redundant, even the entry points and patches in data centers are redundant.
Ideally, the entire network should be designed with full redundancy as a starting point. The network is as simple and straightforward as possible. There is more insight into the risks and there are fewer possible causes in the event of malfunctions. If the configuration is designed this way from the start, end-to-end redundant connectivity can be offered efficiently and is a real option even with smaller budgets.
Techniques such as Software Defined Networking also ensure that unexpected changes in the network are effectively responded to. It offers advantages in the management of the network and reduces the risk of errors due to human actions. In addition, smarter algorithms can be used for problem detection and better routing. For example, routing based on the performance of the route.
The DCspine platform has been designed according to these principles. We believe that online service providers in particular should not settle for less: it is strategically important for them to always be able to deliver their services.
Do you want to know how the DCspine platform works, or do you want to experience for yourself how simple it is? Please contact us or try out the portal free of charge.