1. An agent retrieves PM data from VFNO/NSSMF, which is in China Mobile format (see中国移动5GC网元统计数据需求规范4.0), frequency is about 15 mins/times. Each type has many metrics, for example, AMF PM can have more than 100 metrics.
2. The agent converts the data into VES format, and submits it to DCAE VES collector.
3. VES collector publishes the VES event to DMaaP.
4. DataLake picks up the VES event, and saves it to big data DBs. Since MongoDB will be supported in F release and it is also supported by DataLAke, data is saved to MongoDB. 
5. Data Exposure MS, which is a new component in DataLake, exposes the data in MongoDB to external by a set of REST API. Other components query this API through ONAP EXT-API. Each exposure is defined via a set of configurations, which contain url, from which DB to retrieve data, query sql, response body template, etc. Such config is defined in SDC (TBD). 
6. UUI query MongoDB request data from Data Exposure MS via EXT-API for KPI visualization.
7. For a closed loop control (which is a stretch goal in this release, and the requirement is to be panned out.), TCA subscribes to VES events and triggers alerts based on threshold settings, for downstream components to consume.

 Attached flow-v9.PNG is the sequence diagram, while exposure-ms.PNG is the flow diagram of Data Exposure MS.