Controlling the Execution Mode with Environment Variables and Performance Statement Options

You control the execution mode by using environment variables or by specifying options in the PERFORMANCE statement in high-performance analytical procedures, or by a combination of these methods.

The important environment variables follow:

  • grid host identifies the domain name system (DNS) or IP address of the appliance node to which the SAS High-Performance Server Distributed Mode software connects to run in distributed mode.

  • installation location identifies the directory where the SAS High-Performance Server Distributed Mode software is installed on the appliance.

  • data server identifies the database server on Teradata appliances as defined in the hosts file on the client. This data server is the same entry that you usually specify in the SERVER= entry of a LIBNAME statement for Teradata. For more information about specifying LIBNAME statements for Teradata and other engines, see the DBMS-specific section of SAS/ACCESS for Relational Databases: Reference for your engine.

  • grid mode specifies whether the high-performance analytical procedures execute in symmetric or asymmetric mode. Valid values for this variable are sym for symmetric mode and asym for asymmetric mode. The default is symmetric mode.

You can set an environment variable directly from the SAS program by using the OPTION SET= command. For example, the following statements define three variables for a Teradata appliance (the grid mode is the default symmetric mode):


option set=GRIDHOST      ="hpa.sas.com";
option set=GRIDINSTALLLOC="/opt/TKGrid";
option set=GRIDDATASERVER="myserver";

Alternatively, you can set the parameters in the PERFORMANCE statement in high-performance analytical procedures. For example:


performance host      ="hpa.sas.com"
            install   ="/opt/TKGrid"
            dataserver="myserver";

The following statements define three variables that are needed to run asymmetrically on a computing appliance.


option set=GRIDHOST      ="compute_appliance.sas.com";
option set=GRIDINSTALLLOC="/opt/TKGrid";
option set=GRIDMODE      ="asym";

Alternatively, you can set the parameters in the PERFORMANCE statement in high-performance analytical procedures. For example:


performance host      ="compute_appliance.sas.com"
            install   ="/opt/TKGrid"
            gridmode  ="asym"

A specification in the PERFORMANCE statement overrides a specification of an environment variable without resetting its value. An environment variable that you set in the SAS session by using an OPTION SET= command remains in effect until it is modified or until the SAS session terminates.

Specifying a data server is necessary only on Teradata systems when you do not explicitly set the GRIDMODE environment variable or specify the GRIDMODE= option in the PERFORMANCE statement. The data server specification depends on the entries in the (client) hosts file. The file specifies the server (suffixed by cop and a number) and an IP address. For example:


 myservercop1  33.44.55.66

The key variable that determines whether a high-performance analytical procedure executes in single-machine or distributed mode is the grid host. The installation location and data server are needed to ensure that a connection to the grid host can be made, given that a host is specified. This book assumes that the installation location and data server (if necessary) have been set by your system administrator.

The following sets of SAS statements are functionally equivalent:


proc hpreduce;
   reduce unsupervised x:;
   performance host="hpa.sas.com";
run;

option set=GRIDHOST="hpa.sas.com";
proc hpreduce;
   reduce unsupervised x:;
run;