Scalability On Facebook, Twitter & Co a few things have to be done differently than with traditional web applications where we have a user, a web server, an application server as well as a database server. When the number of requests increases, the problem with scalability almost always resides with the database. Of course a cluster of DB-servers will solve the bottelneck problem for a while. But a DB cluster solution is quite expensive compared to a real scalable solution. When the load cannot be predicted, you have to build the application so that the database won´t become the bottleneck. Therefore, you typically move the application into the cloud. When there are more users, there are more requests. When there are more requests, you can quickly add more servers. When SQL databases reach their scalability limit, you should change your architecture early enough towards NoSQL databases. Only use SQL databases for things that need to be transactional and store and retrieve other data from NoSQL databases. We have done that for example for Laumex where millions of spare parts with a value of many billion euros can be searched quickly. This example shows how to prepare your web application for an unpredictable user request building it on a smart scalable solution. We are Catalysts, software ist our passion.