This depends on APPLICATION requirements. Simple APPS that can run on Linux on a KVM virtual machine can usually port to OpenStack without modifications. For higher-order applications that run on multiple servers and provide their own high AVAILABILITY using common tools like message queues and HAproxy/load balancing, it may be appropriate to architect deployments to reflect OpenStack best-practices for availability (i.e., take in mind fault domains and availability zones to eliminate likely single points of failure), but deeper changes probably won’t be needed. Only in cases where application HA is platform-dependent may deeper reengineering be needed. A GOOD introductory text for modifying applications can be found at: http://pivotal.io/platform-as-a-service/migrating-to-cloud-native-application-architectures-ebook

This depends on application requirements. Simple apps that can run on Linux on a KVM virtual machine can usually port to OpenStack without modifications. For higher-order applications that run on multiple servers and provide their own high availability using common tools like message queues and HAproxy/load balancing, it may be appropriate to architect deployments to reflect OpenStack best-practices for availability (i.e., take in mind fault domains and availability zones to eliminate likely single points of failure), but deeper changes probably won’t be needed. Only in cases where application HA is platform-dependent may deeper reengineering be needed. A good introductory text for modifying applications can be found at: http://pivotal.io/platform-as-a-service/migrating-to-cloud-native-application-architectures-ebook