SAS 9.4 Language Reference, 6th Edition

Previous releases of Base SAS 9.4 support programs written in the SAS DS2

programming language or the SAS Federated SQL language. These languages can take

advantage of threading. Many other SAS products also use threading technology. For

example, the SAS High-Performance Analytics procedures, SAS Stored Processes, and

SAS Embedded Process either execute or generate code that executes in high-

performance distributed computing environments.

What Is Threading Technology in SAS?

Threading technology provides multiple paths of execution within an operating

environment. Each path of execution is called a thread, and each thread can handle a

program task or data transfer. The result is multiple program tasks and data I/O

operations performed at the same time, in parallel. A thread requires a context (like a

register set and a program counter), a segment of code to execute, and some amount of

memory to use in the process. A threading operating environment might have multiple

CPUs but only one core per CPU. Other more high-performance configurations might

include multiple CPUs with multiple cores per CPU and even multiple threads per core.

In situations in which each CPU might execute only one thread at a time, the CPU’s

ability to quickly switch between threads provides near-simultaneous execution.

Threaded execution in SAS software includes one or both of these two general

techniques.

• Threaded I/O means that data (frequently in very high volume) is delivered to an

application in threads so that the application is continually processing, not waiting on

data. In Base SAS and SAS/STAT, several procedures take advantage of threaded

reads. Also Base SAS includes the SPD engine that reads from a data set that is

partitioned to optimize for threaded input to the application. The SAS High-

Performance Analytics procedures require very rapid data delivery. They require

threaded reads from data distributed across a computing cluster to deliver huge

amounts of data to the application (which is also processing on the cluster) and then

write the data in parallel to the data storage appliance. SAS 9.4M5 includes access to

SAS Viya, which supports distributed, in-memory, multithreaded processing. See

“What is SAS Cloud Analytic Services?” on page 403 for more information about

SAS Cloud Analytic Services with SAS Viya.

• Threaded application processing means that the application itself is structured to

perform certain tasks in parallel on multiple-CPU machines. Threaded application

processing enables the application to process large amounts of data to be processed

more quickly because multiple threads execute on smaller segments of data.

Applications can be designed to take advantage of machines with multiple CPUs

whether it is a local four-way desktop or a server-class machine. The SAS High-

Performance Analytics Server executes on appliances that distribute both the data

and copies of the application across the appliance nodes so that the data is co-located

with the application processing

With SAS 9.4 and SAS Analytics 12.1, customers can access a wide variety of products

and components that use threading to support ever-increasing amounts of data as well as

computationally intensive algorithms and models. Base SAS and Foundation SAS

threading technologies support all of these.

208 Chapter 13 • Support for Parallel Processing