다음을 통해 공유


Elastic Framework Services I

How many times have you found infrastructures with a huge use of CPU but not Memory? Or Memory but not CPU? it´s difficult to prevent where we need resources at anytime and then we try to prepare our architectures to be scalable, but the question is: How many resources are not in use in each machine? In a resource optimization strategy, we can avoid to use some resources in our machines because others ar currently in use? We have thinked in how to change this non efficient use. Our first point, has been how to get this new functionality without change the current programming elements, and was necessary keep in mind two possible options:

  • New components and services, ready to scale in a elastic model
  • Existing component and services, not ready to scale in a elastic model

Our second challenge was, how to built this new model and don´t make complex the builting and debugging process to developers, testers, even the Architects will not be affected in their designs. The model is oriented to be all possible transparent for them. In the next posts, we will share more about this new model: EFS