Performance measurement of nova services

Today, scaling problems in nova are generally solved by adding resources (either scale out or scale up) to the controller. For any product, gaining insight into hot code paths, memory utilization is equally important. Such insights usually lead to optimization to codebase resulting in better resource utilization and is a mean to achieving scale.
In context of nova, profiling information can include
1. for each request, breakup of time spent in each service
2. memory utilization of each service per request
3. objects allocated by (mainly nova-conductor) depending on inventory size
4. time spent in DB queries made and in processing DB information.

There can be many more.