Added parse.rst and mrtask.rst.

h2o-docs/source/developuser/mrtask.rst

@@ -0,0 +1,13 @@
+Job/MRTask/FJTask Overview
+==========================
+
+A GLM job gets broken down into MRTask2 tasks, and subsequently into
+Fork/Join tasks.  FJ tasks are spread across the cluster in a
+logarithmic tree fashion with computation performed at the leaves and
+results rolled up to the top.
+
+.. image:: PngGen/pictures/GLMAlgoMem.png
+   :width: 90 %
+
+|
+|

h2o-docs/source/developuser/parse.rst

@@ -0,0 +1,29 @@
+Parse Overview
+==============
+
+(Since HDFS is very popular, this example illustrates that data
+source.  H2O does support other sources of data, however.)
+
+The parse process typically moves data as much as twice during
+ingestion.
+
+The first movement of the data occurs when the data is read from disk
+(an f-chunk in the diagram below) and copied across the network to the
+H2O node that requested the particular piece of data from the
+filesystem.
+
+.. image:: PngGen/pictures/DataIngestion.png
+   :width: 90 %
+
+|
+|
+
+The data is then moved a second time from the H2O node where the raw
+data gets parsed to the H2O node where the compressed data will reside
+in a Fluid Vector chunk (a p-chunk in the diagram below).
+
+.. image:: PngGen/pictures/Parse.png
+   :width: 90 %
+
+|
+|

h2o-docs/source/developuser/top_developer.rst

@@ -27,3 +27,5 @@ H2O Software Architecture
 
    h2o_stack
    how_r_interacts_with_h2o
+   parse
+   mrtask