The following list identifies
some usage notes:
-
-
You cannot replace or overwrite
tables in memory. You must unload the table and then load the new
table.
-
You cannot use the APPEND procedure.
However, you can use an APPEND= data set option to achieve the same
result.
-
Loading tables into memory from
HDFS is performed with the LASR procedure. You cannot load tables into memory from HDFS
with the engine.
-
The engine guarantees the data
order for a particular configuration of worker nodes. If you load
a table in to a server and you retrieve the data three times, the
order of the data is the same. However, if you start another server
and load the same table into a different number of worker nodes, then
the order in which you retrieve the data is different. However, it
is reproducible within fetches from a single server.
-
Any order-dependent operation,
such as the LAG or DIF functions, cannot rely on stability of results
beyond that which can be guaranteed by the distribution model of the
data.
When the engine is used
to access a SAS LASR Analytic Server
that is on the same machine, a network socket connection is used.
This applies to Windows and Linux operating environments. However,
when you use the engine to connect to a server on a different host,
the connection is based on passwordless SSH. As a result, the engine
can connect to only remote hosts that use Linux.