You can assign keys based on the time rows are added or a unique ID algorithm. You can also use an existing column in the dataset as the key or a combination of columns. For example:
See
<storeto> for detailed information on the methods to derive keys.