Cloud computing has become increasingly important in the area of information technology over the past few years and the trend is expected to continue. One of the most widespread systems is the open source based OpenStack platform, which is used by many industry players. Cloud provide services that include automatic scaling, which has some constraints on scaling-in, which may be a problem for some customers. The source of this problem is that when the virtual machines are shut down, the status of the current service is ignored from the machine which is being shut down. The problem is solved by a system where the component which will shut down the virtual machine has access to the measurement data from the affected machines. This way the system can make a smarter decision instead of the original one which always stopped the oldest ones. For implementation, I installed my own OpenStack environment using two computers of the university lab. After configuring all the required components, I created an environment descriptor template based on the orchestration module could create the resources which defined in it. With this I was able to provide load-based up-and-down scaling. After that, I designed a functional supplement for the development of sculpture and implemented it. I have verified the successful implementation by functional testing and presented its use possibilities. Finally, I gave an insight into the possibilities to continue the work.