RSS Matters

The All-encompassing SAS 8 (2/2)*

By Dr.Karl Ho, Research and Statistical Support Services Manager

In the previous article, I introduced some of the new features in SAS versions 8 and 8.1. In this article I decided to cover some very basic but more technical parts of the new SAS design and its upcoming development directions.

Beginning in version 8, a product of the Nashville project that transitioned SAS into the real 32-bit computing environment, the SAS software evolved into a data system that in itself performs a multitude of statistical analyses and data functions. A number of add-on modules were built into the system, rendering the software one of the largest programs, which takes a minimum of 850 megabytes in typical installation.  With the documentation and GIS files installed, this number can amount up to 2 gigabytes. 

SAS 8 is engineered for 32-bit operating environments, namely, Windows 2000 although it works fine in Windows 98 and Windows NT. It taps into the server capabilities of the operating system to facilitate the client-server task sharing operations, and allow multiple users to share server or servers. While Windows 2000 incorporates a number of Windows NT designs/features, such as Terminal Services, it provides more room for SAS to further its client-server development designed for division of labor among computers. 

For instance, the new MP CONNECT procedure in SAS 8 performs parallel multiprocessing that can be up to six times faster than serial execution (Garner 2000). Imagine running SAS on the Data Center Server version of Windows 2000 that accommodates up to 32 CPU's in handling 64 GB file operations--which will not be uncommon in financial database operations-- as compared to a scaled down version of Windows 2000 Pro utilizing 2 CPUs with 2 GB RAM each. It is just like juggling multiple US Census files (exceeding 2 GBs) in a few minutes! Plus, can you find other software (not a programmer to custom develop codes) to perform the job?

When hardware is in the picture, SAS 8's edge widens, as it exploits Pentium III's Streaming SIMD instructions to facilitate faster completion of certain types of data operations such as data retrieval and floating point calculation. According to SAS' technical report, the new design in SAS 8 will have the processor pre-fetch data before they are actually needed for calculations. As a result, data processing can be as quick as the processor can run since data retrieval time is reduced to minimal (Mehler 2000).   Remember, we are not talking about 50 or 100 megabytes of data. When gigabytes of data are involved, it can be translated into a matter of tens of minutes and even hours! With Windows 2000 NTFS, the system becomes genuinely capable of individual files exceeding 2 GB in size.

Future development

SAS 8.1 is geared toward taking advantage of the new Intel-based microprocessors and Microsoft's Windows 2000 operating environment.Within two years, the new generation of PC-based processor from Intel, the 64-bit Itanium, will replace the 32-bit Pentium and prepare desktop computing for the genuine parallel processing in the future server-to-server environment. With 64-bit addressing, at our finger tips is a machine that can have the number of memory addresses with digits our fingers can't count (yes, plus our toes)**. SAS' edge over other data analysis and statistical program may widen even more then when their current R&D on parallel processing pays off. On the horizon, we are anticipating desktop applications such as SAS that capitalizes the hardware data capability to share computation with remote servers or centralized data warehouses. In fact, servers will abound then (who do not have one!) and throughput will be our main concern about computation. \

References

Garner, Cheryl. 2000. Multiprocessing with Version 8 of the SAS System SAS SUGI Paper 16-25

Mehler, Gary. 2000. Taking Advantage of the SAS System on the Windows Platform.  SAS SUGI Paper 280-25

Robert Ray. 2000. Version 8 Base SAS® Performance: How Does It Stack-Up? SAS SUGI Paper 9-25

What's New in Data Analysis on SAS Research and Development communities Web (http://www.sas.com/rnd/app/da/danew.html)

* In the next article, I will go back to more new SAS procedures and modules such as IML Workshop and Partial Least Square

** To be exact, it is 2⁶⁴, or 18,446,744,073,709,551,616, bytes.