A multinational's website makes use of two login types: name and IP. Logging in using IP caused problems: whenever over 20 concurrent users logged in, the website went down. Our goal became to improve the site's performance to over 500 concurrent users. During Analysis, the thirds step of our procedure, possible corrections in installation and configuration came up. However, even greater improvements showed up during analysis of indexes. After adding new indexes and changing some of the existing ones, the website is accessible to more than 600 concurrent users. This way, our goal was reached.
A company in marketing software encountered an issue being unable to process all messages a larger sport manufacturer posted. Analysis showed where the bottleneck was situated: all messages, processed and to be processed passed one and the same table. As messages were added faster than they could be processed, the queue only grew. Our goal: getting the system to process incoming messages faster than they came in, to keep the queue as small as possible. Our solution: partitioning this table, making distinction on the message's status ('processed' and 'to be processed'). Since our proposition has been implemented, the system is able to process 40.000 messages an hour, instead of a maximum of 2.000 messages in the past. Once again our goal was reached.