This section describes key concepts associated with Dremio datasets.
In Dremio, datasets are either physical or virtual datasets.
Physical datasets (PDS) are represented by these icons in the Dremio UI. The first represents a PDS created from a folder of files, and the second represents a PDS created from a single file:
They are stored within the data sources. Physical datasets cannot be modified by Dremio.
Partition information is available for the columns of each dataset, just hover over the icon to see the information icon. To see the partition information, click on PDS icon.
Virtual datasets (VDS) are represented by this icon in the Dremio UI:
They are derived from physical datasets or other virtual datasets. Virtual datasets are defined by the steps necessary for their creation, including transformations, filters, joins, and other modifications. Because virtual datasets do not make a copy of the data, they use very little space, and they always reflect the current state of the physical datasets they are derrived from.
Suppose we have a collection on a MongoDB source called sales. Inside Dremio, this collection is represented as a physical dataset under ‘sales.’ We can open this dataset within Dremio, and then save it as a virtual dataset called ‘salesRaw.’ Later on we can derive another virtual dataset called ‘salesNY’ from ‘salesRaw’ by excluding all data that doesn’t originate from the state of New York. ‘salesRaw’ and ‘salesNY’ can each be queried and will return different results, but they are both based on the same underlying physical dataset.
Spaces are where virtual datasets are saved. Spaces provide a way to group datasets by a common theme such as a project, department, or geographic region. Each user also has a default Home Space.
For instance, a list of spaces for an online retailer might look like:
You can use folders to provide a deeper layer of organization to spaces. Folders can contain other folders.
Paths are a dot-separated list that indicates the location of a dataset, starting with the name of the source or space in which that dataset resides, followed by any folders or data source structures, and ending in the name of the dataset. Here are a few examples of what dataset paths look like in Dremio:
Transactions is a space,
regions is a folder, and
salesNY is a virtual dataset. In the second example,
Web Traffic is a file system data source,
october is a directory on that file system, and
visits is a sub-directory with a group of files in a common structure.
SQL queries always reference data sets using their full path, for example
SELECT * FROM "web traffic".october.visits.