The Doppler Quarterly Spring 2019 - Page 66

The Right Data in the Right Place So far, we have assumed that we are going to be replicating our entire data estate in all the places where we have appli- cations. In some scenarios, this may be the best (and only) option. However, with careful planning and potentially some changes to our applications and their interactions, we can take a more efficient and effective approach. Many organizations have application sets that can work independently with regard to their associated data. As we move closer and closer to a microservices architecture, we want that independence in order to achieve the agility and functional isolation that microservices offer. But even prior to achieving microservice nirvana, we probably have appli- cations that only need to update other data sets infre- quently and with small amounts of additions and updates. Consider a scenario where we have applications in one pub- lic cloud which primarily do data analytics for operational capabilities like machine learning. It is entirely probable that the majority of the data required will stay resident within that cloud provider, with new data being added from exter- nal sources (ingress). The results from the machine learning models could be fed to other clouds or back to the on-prem- ises environment in small chunks, avoiding large costs and other factors such as latency. Public Internet Premise Data Center Figure 6: The right data in the right place, multicloud approach 64 | THE DOPPLER | SPRING 2019 User