SAS/ACCESS Engines
SAS/ACCESS engines are LIBNAME engines that provide Read, Write, and Update
access to more than 60 relational and nonrelational databases, PC files, data warehouse
appliances, and distributed file systems. These engines are not part of Base SAS but they
depend on Base SAS. They are licensed separately or are included in many product
bundles such as SAS BI Server or SAS Activity-Based Management. Many bundles
offer the customer a choice of two out of the many SAS/ACCESS engines available.
SAS/ACCESS engines enable SAS programs to connect to a DBMS as if it were a SAS
data set. This takes advantage of performance-related DBMS features and benefits
including bulk load support, temporary table support, and native SQL support with
Explicit Pass-Through. If the DBMS is a parallel server, the engine accesses the DBMS
data in parallel by using multiple threads to connect to the DBMS server. If your SAS
program is executing a thread-enabled SAS procedure with these SAS/ACCESS engines,
even greater gains in performance are likely.
In SAS/ACCESS, threaded reads partition the result set across multiple threads. Unlike
threaded processing in Base SAS procedures, threaded reads in SAS/ACCESS are not
dependent on the number of processors on a machine. Instead, the result set is retrieved
on multiple connections between SAS and the DBMS. SAS causes the DBMS to
partition the result set by appending a WHERE clause to the SQL statement. When this
happens, a single SQL statement becomes multiple SQL statements, one on each thread.
The DBMS reads the partitions one per thread also.
The amount of scalability that is provided with the SAS/ACCESS engines depends on
the efficiency of parallelization implemented in the DBMS itself. However,
SAS/ACCESS engines have options available in the LIBNAME statement that enable
tuning of the threaded implementation within the SAS/ACCESS engines. The options
that control threaded reads in SAS/ACCESS are DBSLICE, DBSLICEPARM,
THREADS|NOTHREADS, and whether BY, OBS, or KEY options are used in a PROC
or DATA step. Refer to the SAS/ACCESS for Relational Databases documentation for
more information.
SAS Scalable Performance Data Server
SAS Scalable Performance Data Server is a multi-user parallel-processing data server
with a comprehensive security infrastructure, backup and restore utilities, and
administrative and tuning options. SPD Server supports native SQL, FedSQL, and DS2
languages. Options specific to the SPD Server control threaded processing. The
NOTHREADS and CPUCOUNT options have no effect. See the SAS Scalable
Performance Data Server: Administrator’s Guide and SAS Scalable Performance Data
Server: User’s Guide for information.
SAS Intelligence Platform
The SAS Intelligence Platform is an infrastructure for creating, managing, and
distributing enterprise intelligence. This infrastructure supports SAS solutions for
industries such as financial services, life sciences, health care, retail, and manufacturing.
212 Chapter 13 • Support for Parallel Processing
The SAS Intelligence Platform is not part of Base SAS but instead it relies on SAS
Foundation, which includes Base SAS and these components:
• SAS Management Console for defining metadata
• SAS Business Intelligence Server and SAS Data Integration Server technologies
• SAS Integration Technologies, which provides the following:
• the SAS servers and supporting services such as SAS Stored Process Servers
• application messaging
• SAS BI Web Services
• publishing framework
• SAS Foundation Services
The SAS Intelligence Platform Overview discusses individual components and references
a wide spectrum of related SAS Intelligence Platform documentation.
These are the SAS servers in the Intelligence Platform:
• SAS Workspace Server
• SAS Stored Process Server
• SAS Pooled Workspace Server
• SAS OLAP Server
• SAS Metadata Server
Each server is initiated with a pool of active threads. These threads are controlled by the
server and are used by server processes (for example, handling incoming requests). If the
NOTHREADS and CPUCOUNT options are specified, they are ignored, except during
the execution of submitted code that includes a SAS procedure that honors these options.
For the SAS Metadata Server, thread usage is controlled by default settings for the object
server parameters (THREADSMAX and THREADSMIN) and for the metadata server
configuration option, MACACTIVETHREADS. Administrators can override these
settings in order to fine-tune performance. See the SAS Intelligence Platform: System
Administrator’s Guide for details and examples. The THREADMAX and THREADMIN
object server parameters are rarely used for servers other than the SAS Metadata Server.
In the intelligent platform middle tier (which is an infrastructure for web applications),
incoming requests are processed on threads. These threads are defined using the job
execution service. The threads are not constrained by the NOTHREADS or
CPUCOUNT options. Both the number of job queue threads and number of job
execution threads can be specified. Refer to the SAS Intelligence Platform: Middle-Tier
Administration Guide.
SAS MP CONNECT is a part of SAS/CONNECT software that is bundled with the
intelligence platform. It supports parallel processing by establishing a connection
between multiple SAS sessions and enabling each of the sessions to asynchronously
execute tasks in parallel. By establishing connections to processes on the same local
computer, the application can use network resources to process in parallel and coordinate
all the results into the client SAS session. Many SAS processes use multiple processors
on an SMP computer, but they can also be executed on multiple remote single or
multiprocessor computers on a network. Threads are always assumed to be available.
Some SAS High-Performance Analytics products can execute on the SAS Intelligence
Platform if it is configured as an MPP environment. For example, SAS Grid Manager
(discussed in the next section) handles workload management for SAS applications that
execute in SMP configurations. It can also manage applications that are coded for
SAS Intelligence Platform 213
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.