We’re smarter together. Learn from this collection of community knowledge and add your expertise.

SAS data format for Hadoop is open to open source

by SAS Employee SteveSober on ‎03-15-2016 09:49 AM - edited on ‎03-15-2016 09:50 AM by Community Manager (2,293 Views)

In a recent article we learned how to create an SAS Scalable Performance Data Engine (SPD Engine) table on HDFS. In this article we will explore the Hive SerDe for SPD Engine data, which provides read access to the SPD Engine table using Hive via non-SAS clients.


In this example, we will create two uncompressed SPD Engine tables stored on HDFS. We’ll then register those tables to the Hive metastore and join them together to create a Hive target table.


Code to create the two SPD Engine tables:

59        +LIBNAME SPDEHDFS SPDE '/user/sasss1/spde' HDFSHOST=default ACCELWHERE=YES 
PARALLELWRITE=YES PARALLELREAD=YES; NOTE: Libref SPDEHDFS was successfully assigned as follows: Engine: SPDE Physical Name: /user/sasss1/spde/ 66 +DATA SPDEHDFS.PLAYERS; 67 + KEEP name team; 68 + SET sashelp.baseball; 69 +RUN; NOTE: There were 322 observations read from the data set SASHELP.BASEBALL. NOTE: The data set SPDEHDFS.PLAYERS has 322 observations and 2 variables. NOTE: DATA statement used (Total process time): real time 1.75 seconds cpu time 0.16 seconds 70 + 71 +DATA SPDEHDFS.STATS; 72 + KEEP name nhits nruns; 73 + SET sashelp.baseball; 74 +RUN; NOTE: There were 322 observations read from the data set SASHELP.BASEBALL. NOTE: The data set SPDEHDFS.STATS has 322 observations and 3 variables. NOTE: DATA statement used (Total process time): real time 1.38 seconds cpu time 0.15 seconds


For SAS Administrators:


Hive SerDe for SPD Engine Data

In the third maintenance release for SAS 9.4, SAS provides a custom Hive SerDe for SPD Engine Data that is stored in HDFS. The SerDe makes the data available for applications outside of SAS to query.


The following are required to access SPD Engine tables using the SPD Engine SerDe:

  • You must deploy SAS Foundation using the SAS Deployment Wizard. Select SAS Hive SerDe for SPDE Data.

You must be running a supported Hadoop distribution that includes Hive 0.13:

  • Cloudera CDH 5.2 or later
  • Hortonworks HDP 2.1 or later
  • MapR 4.0.2 or later

The following table features are not supported:

  • Compressed or encrypted tables
  • Tables with SAS informats
  • Tables that have user-defined formats
  • Password-protected tables
  • Tables owned by the SAS Scalable Performance Data Server

In addition, the following processing functionality is not supported by the SerDe and requires processing by the SPD Engine:

  • Write, Update, and Append operations
  • If preserving observation order is required.


On the Hadoop head node

To register the SPD Engine tables to the Hive metastore we need to append to the HADOOP_CLASSPATH the SAS_HADOOP_CONFIG_PATH, as well as the directory paths for the Hive JARs and XMLs files:


export HADOOP_CLASSPATH=$HADOOP_CLASSPATH:$SAS_HADOOP_CONFIG_PATH:opt/cloudera/parcels/CDH/lib/hive/lib/*:/opt/cloudera/paels/CDH/lib/hadoop-mapreduce/lib/*


In the following two examples, we are registering the SPD Engine tables “players and stats” to the Hive “sasss1” database. The following commands executes the SerDe JAR files and registers the SPD Engine tables to the Hive metastore:




hadoop jar /opt/cloudera/parcels/CDH/lib/hive/lib/sas.HiveSerdeSPDE.jar com.sas.hadoop.serde.sp.hive.MetastoreRegistration -table players -mdflocation /user/sasss1/spde -database sasss1

hadoop jar /opt/cloudera/parcels/CDH/lib/hive/lib/sas.HiveSerdeSPDE.jar com.sas.hadoop.serde.sp.hive.MetastoreRegistration -table stats  -mdflocation /user/sasss1/spde -database sasss1

When registering the tables to the Hive metastore the following options are also supported:


-database database_name to specifies a Hive metastore database if you are not using the default database.

-mdflocation is the path on HDFS that we used on our SPD Engine LIBNAME.

-renametable table-name to assigns a different name to the table stored in HDFS.

-owner owner-name that assigns an owner name. The default owner name is Anonymous.


For SAS Programmers:

In this example we will create a new Hive table by joining two SPD Engine tables in the MapReduce framework by using the two registered table in the Hive metastore:

58        +/* HIVE LIBREF */
HOST="Hadoop.Head.Node" DATABASE=sasss1 USER=sasss1; NOTE: Libref HIVE was successfully assigned as follows: Engine: HADOOP Physical Name: jdbc:hive2://Hadoop,head.node.com:10000/sasss1 102 +PROC SQL; 103 + CREATE TABLE HIVE.PLAYER_STATS AS 104 + SELECT t1.Name, 105 + t1.Team, 106 + t2.nHits, 107 + t2.nRuns 108 + FROM HIVE.PLAYERS t1 109 + INNER JOIN HIVE.STATS t2 ON (t1.Name = t2.Name); NOTE: SAS variable labels, formats, and lengths are not written to DBMS tables. NOTE: Table HIVE.PLAYER_STATS created, with 322 rows and 4 columns. HADOOP_23: Executed: on connection 4 DROP TABLE sasdata_17_43_43_325_00001 Summary Statistics for HADOOP are: Total SQL execution seconds were: 0.137603 Total SQL prepare seconds were: 16.867544 Total SQL describe seconds were: 0.002308 Total seconds used by the HADOOP ACCESS engine were 19.764125 110 +QUIT; NOTE: PROCEDURE SQL used (Total process time): real time 20.61 seconds cpu time 0.17 seconds


As we can see from Figure 1, the above join is using MapReduce to accomplish the join:

Figure1.pngFigure 1. Cloudera monitor validating the join is being processed via MapReduce



Notice in Figure 2 the SAS Server that submitted this code is not consuming resource:


Figure 2. Top command, notice SAS and JPROXY are not consuming resources


Non-SAS clients

Once the SPD Engine tables are register to the Hive meatastore any non-SAS tool used to process Hive data can also process (read only) the SPD Engine data using HiveQL.

Your turn
Sign In!

Want to write an article? Sign in with your profile.

Looking for the Ask the Expert series? Find it in its new home: communities.sas.com/askexpert.