diff --git a/docs/getting_started/file_io.rst b/docs/getting_started/file_io.rst
new file mode 100644
index 00000000..6ca96dbd
--- /dev/null
+++ b/docs/getting_started/file_io.rst
@@ -0,0 +1,84 @@
+.. _file_io:
+
+File I/O
+========
+
+This page covers reading and writing raster files with rioxarray. For the basics of
+opening files, see :ref:`getting_started`.
+
+Reading Files
+-------------
+
+Cloud storage
+~~~~~~~~~~~~~~
+
+Paths are passed through to rasterio/GDAL, so you can also read files directly
+from HTTP(S) URLs and cloud storage (e.g. ``s3://``). See :ref:`gdal_env` for
+configuring GDAL when working with cloud-stored data.
+
+.. code-block:: python
+
+    import xarray
+
+    xds = xarray.open_dataset("s3://bucket/my.tif", engine="rasterio")
+
+
+Writing Files
+-------------
+
+GeoTIFF
+~~~~~~~~
+
+Use ``rio.to_raster`` on an ``xarray.DataArray`` or ``xarray.Dataset`` to write a raster.
+The driver is detected from the ``.tif``/``.tiff`` extension. Additional keyword arguments
+are passed to rasterio as GDAL creation options.
+
+.. code-block:: python
+
+    import rioxarray
+
+    xds = rioxarray.open_rasterio("my.tif")
+    xds.rio.to_raster("out.tif")
+    xds.rio.to_raster("out_lzw.tif", compress="LZW")
+
+Cloud Optimized GeoTIFF (COG)
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+To write a COG, use GDAL's ``COG`` driver:
+
+.. code-block:: python
+
+    xds.rio.to_raster("out_cog.tif", driver="COG")
+
+For more examples, including writing large rasters with ``windowed=True`` or with dask,
+see :doc:`../examples/convert_to_raster`, :doc:`../examples/COG`,
+and :doc:`../examples/dask_read_write`.
+
+Cloud storage
+~~~~~~~~~~~~~~
+
+To write directly to cloud storage, use a
+`GDAL virtual file system <https://gdal.org/en/stable/user/virtual_file_systems.html>`__
+path such as ``/vsis3/`` (``/vsigs/`` and ``/vsiaz/`` work similarly).
+These file systems only support sequential writes, so set
+``CPL_VSIL_USE_TEMP_FILE_FOR_RANDOM_WRITE`` to have GDAL write to a local temporary file
+and upload it when the file is closed:
+
+.. code-block:: python
+
+    import rasterio
+
+    with rasterio.Env(CPL_VSIL_USE_TEMP_FILE_FOR_RANDOM_WRITE="YES"):
+        xds.rio.to_raster("/vsis3/bucket/out.tif", driver="COG")
+
+Alternatively, ``rio.to_raster`` accepts a file-like object, such as an ``io.BytesIO``
+buffer or a file opened with `fsspec <https://filesystem-spec.readthedocs.io>`__
+(writing to ``s3://`` with fsspec requires `s3fs <https://s3fs.readthedocs.io>`__).
+The ``driver`` must be set explicitly because it cannot be detected from a file extension.
+
+.. code-block:: python
+
+    import fsspec
+
+    with fsspec.open("s3://bucket/out.tif", "wb") as fobj:
+        xds.rio.to_raster(fobj, driver="COG")
diff --git a/docs/getting_started/getting_started.rst b/docs/getting_started/getting_started.rst
index 01d62cb0..f234372d 100644
--- a/docs/getting_started/getting_started.rst
+++ b/docs/getting_started/getting_started.rst
@@ -73,6 +73,10 @@ Why use :func:`rioxarray.open_rasterio` instead of `xarray.open_rasterio`?
 7. `xarray.open_rasterio` is deprecated (since v0.20.0)
 
 
+See :ref:`file_io` for reading from cloud storage and for writing files
+(GeoTIFF, Cloud Optimized GeoTIFF, and cloud storage).
+
+
 Introductory Information
 --------------------------
 
@@ -80,6 +84,7 @@ Introductory Information
    :maxdepth: 1
 
    switching_from_rasterio
+   file_io
    crs_management.ipynb
    nodata_management.ipynb
    manage_information_loss.ipynb
diff --git a/docs/history.rst b/docs/history.rst
index c07fc71b..7509a949 100644
--- a/docs/history.rst
+++ b/docs/history.rst
@@ -10,6 +10,8 @@ Latest
   coordinates, which also fixes the output size of ``rio.reproject`` with
   ``src_geoloc_array`` (issue #848)
 
+- DOC: Add file I/O page with writing files section (issue #584)
+
 0.23.0
 ------
 - ENH: Add write support for Zarr spatial and proj conventions