Introduction

Starting something new, whether a personal or professional project, can be intimidating. Yet, it is often a very rewarding process. If you are hesitant to start programming with SAS^©® Cloud Analytic Services (CAS), know that the benefits are considerable.

CAS is the analytical engine at the heart of SAS^© Viya^™. It is a server that provides a cloud-based execution environment for data management and analysis. Its strength lies in its ability to process data in-memory and in a distributed manner across multiple machines, using multithreaded processing. This allows for extremely high-performance analysis on very large tables.

This article aims to guide you through your first steps with CAS, covering the entire process from establishing a session to analyzing the data.

The CAS Architecture in Brief

The CAS server can run on a single machine (SMP) or in a distributed mode on multiple machines (MPP). The distributed architecture includes:

A Controller node: It communicates with client applications and directs the work.
Worker nodes: They perform calculations and analysis on the data rows stored in memory on their respective node.

Get Started! A Beginner's Guide to Programming in the SAS® Cloud Analytic Services (CAS) Environment -

The Key Advantages of CAS

Here is a summary of the main advantages of using the CAS environment:

Table 1: Benefits of the CAS Environment

Benefit	Description
Processing Speed	The distributed architecture and in-memory processing allow for extremely fast analyses on large datasets.
Fault Tolerance	If a worker node stops responding, the controller redirects the work to another node using a copy of the data, ensuring continuity.
Scalability	It is possible to horizontally add worker nodes to distribute the load and improve processing times.
Open Codebase	Although this article focuses on SAS^©, CAS also supports languages like Python, Java, Lua, and R.

Prerequisite: Using SAS^©® Studio

For SAS^© programmers, SAS^© Studio is the recommended interface for interacting with CAS. A web interface accessible from any device, it offers effective tools like predefined code snippets for common tasks (e.g., listing CAS session options).

The CAS Programming Workflow

The typical CAS programming process follows these steps:

Establish a CAS session.
Allocate a CAS library (caslib).
Load data into CAS.
Manipulate data (e.g., with the DATA step).
Execute actions with PROC CAS.
Analyze the data.

Step 1: Establish a CAS Session

The first mandatory step is to initiate a CAS session. This is the communication channel between your client (SAS^© Studio) and the CAS server. The session manages your authentication, resource allocation, and ensures process isolation.

The basic syntax is very simple:

1	cas casauto;

This statement starts a session named CASAUTO with default properties. If the connection is successful, the SAS^© log will display information about the port, the session UUID, the user, the default active caslib, and the number of worker nodes used.

Step 2: Allocate a CAS Library (Caslib)

"Caslibs" are the storage spaces for your in-memory data in CAS. They also contain the access controls for the data.

Types of caslibs: Personal (accessible only by you), Predefined (global, managed by the administrator), or Manually Added.
Scope: "Session" (visible only in your current session) or "Global" (shared between sessions).

To list the properties of your default active caslib (often CASUSER), use:

1	caslib casuser list;

Important: The Link between Traditional SAS^© and CAS The name of a caslib is not a standard SAS^© "libref". To use traditional SAS^© procedures or the DATA step with CAS tables, you must associate a libref with the caslib via the LIBNAME statement using the CAS engine:

1	LIBNAME mycas cas caslib=casuser;

Step 3: Load Data into CAS

CAS offers great flexibility for loading data (SAS^© files, CSV, Excel, databases...). Here are three common methods:

Table 2: Methods for Loading Data into CAS

Method	Description	Simplified Code Example
DATA Step	Ideal for experienced SAS^© programmers. Allows loading existing SAS^© datasets.	`data mycas.cars; set sashelp.cars; run;`
PROC CASUTIL	Utility procedure dedicated to managing CAS files and tables. Similar to PROC IMPORT for external files.	`proc casutil; load file='/path/cars.xls' casout='cars2' importoptions=(filetype='xls'); quit;`
PROC CAS (loadTable Action)	Uses the CASL language to execute server actions. A powerful and scriptable method.	`proc cas; table.loadtable / path='cars.csv' casout={name='cars3'} ...; quit;`

Step 4: Manipulate Data with the DATA Step

The good news for SAS^© users is that the trusty DATA step works in CAS.

The Major Difference: Parallel Processing In CAS, the DATA step runs in a distributed manner. Very large datasets are divided among the available "threads" on the different machines. The DATA step code is copied and executed simultaneously on each thread, processing only the portion of data local to that thread.

Example of Parallel Processing: In the example below, we add transaction fees to a large banking table. Using the automatic variable _THREADID_ in the log shows that the code is running on multiple different threads (e.g., 4 threads).

1	DATA mycas.updated_transaction_history;
2	SET mycas.transaction_history;
3	/* Logique pour ajouter des frais selon l'année */
4	IF year(transaction_dt)=2013 THEN fee=1;
5	/* ... autres années ... */
6	new_transaction_amt=transaction_amt+fee;
7	put _threadid_=; /* Affiche le numéro du thread dans le journal */
8	RUN;

The log will display messages like NOTE: Running DATA step in Cloud Analytic Services. and multiple lines for _THREADID_=1, _THREADID_=2, etc.

Caution with BY-group Processing: Due to data distribution, the order of BY-group processing is not guaranteed in CAS, unlike in classic SAS^©. Groups are distributed across nodes, and each node processes its groups independently to speed up the process.

Step 5: Use PROC CAS to Execute Actions

PROC CAS is the interface for executing the CAS Language (CASL). CASL interacts with the server via "actions". Actions are requests for specific tasks (table management, analyses, etc.), grouped into "action sets".

Here are examples of actions from the table action set for managing data:

Check if a table exists and retrieve info:

1	PROC CAS;
2	SESSION casauto;
3	/* Vérifier si la table existe */
4	TABLE.tableexists RESULT=r / caslib='casuser' name='updated_transaction_history';
5
6	IF (r.exists) THEN DO;
7	/* Obtenir les infos de la table */
8	TABLE.tableinfo / caslib='casuser' name='updated_transaction_history';
9	/* Récupérer (fetch) un échantillon de lignes */
10	TABLE.fetch / TABLE={caslib='casuser', name='updated_transaction_history'} from=1 to=20;
11	END;
12	QUIT;

Promote a table (make it global): Tables created often have "session" scope. The promote action allows them to be made available to other users.

1	PROC CAS;
2	SESSION casauto;
3	TABLE.promote / caslib='casuser' name='updated_transaction_history' targetlib='casuser';
4	QUIT;

Step 6: Analyze the Data

Method A: Use a CAS Action (e.g., simple.freq) The simple action set provides basic analytical functions.

1	PROC CAS;
2	SESSION casauto;
3	/* Distribution de fréquence du statut de transaction groupé par année */
4	SIMPLE.freq /
5	inputs={'transaction_status'}
6	TABLE={caslib='casuser', name='updated_transaction_history', groupby={name='year'}};
7	QUIT;

Method B: Use a SAS^© Viya^™ Procedure (e.g., PROC MDSUMMARY) SAS^© Viya^™ offers many procedures (statistics, data mining, machine learning) that run in CAS. PROC MDSUMMARY is a powerful statistical procedure for calculating descriptive statistics.

1	/* Préparation des données : ajout d'une colonne mois */
2	DATA mycas.updated_transaction_history2;
3	SET mycas.updated_transaction_history;
4	month=put(transaction_dt,monname8.);
5	RUN;
6
7	/* Calcul des sommes par année et mois */
8	PROC MDSUMMARY DATA=mycas.updated_transaction_history2(where=(fee ne 0));
9	var fee;
10	groupby year month / out=mycas.summary_transaction_history;
11	RUN;
12
13	/* Affichage des résultats (car MDSUMMARY ne produit qu'une table de sortie) */
14	PROC PRINT DATA=mycas.summary_transaction_history label;
15	title 'Résumé de l\'historique des transactions';
16	var year month _sum_;
17	label _sum_='Total collecté ($)';
18	FORMAT _sum_ dollar8.;
19	RUN;

SAS^© Cloud Analytic Services (CAS) offers a powerful, distributed, and in-memory environment to accelerate your analyses. By following the steps of connecting, loading data via caslibs, and using familiar tools like the DATA step or new ones like PROC CAS, you can start harnessing the power of SAS^© Viya^™ today. Don't be afraid to try something new!

Important Disclaimer

The codes and examples provided on WeAreCAS.eu are for educational purposes. It is imperative not to blindly copy-paste them into your production environments. The best approach is to understand the logic before applying it. We strongly recommend testing these scripts in a test environment (Sandbox/Dev). WeAreCAS accepts no responsibility for any impact or data loss on your systems.

Back to article list

Get Started! A Beginner's Guide to Programming in the SAS® Cloud Analytic Services (CAS) Environment

Difficulty Level

Published on : 25/08/2022

Expert Advice

Introduction

The CAS Architecture in Brief

The Key Advantages of CAS

Prerequisite: Using SAS^©® Studio

The CAS Programming Workflow

Step 1: Establish a CAS Session

Step 2: Allocate a CAS Library (Caslib)

Step 3: Load Data into CAS

Step 4: Manipulate Data with the DATA Step

Step 5: Use PROC CAS to Execute Actions

Step 6: Analyze the Data

Important Disclaimer

Difficulty Level

Published on : 25/08/2022

Expert Advice

Introduction

The CAS Architecture in Brief

The Key Advantages of CAS

Prerequisite: Using SAS©® Studio

The CAS Programming Workflow

Step 1: Establish a CAS Session

Step 2: Allocate a CAS Library (Caslib)

Step 3: Load Data into CAS

Step 4: Manipulate Data with the DATA Step

Step 5: Use PROC CAS to Execute Actions

Step 6: Analyze the Data

Important Disclaimer

Prerequisite: Using SAS^©® Studio