SAS/CONNECT User's Guide

Considerations for Using CEDA

It is important to understand that CEDA does not replace data transfer services or remote library services because of the following restrictions:

CEDA fully supports only SAS data sets. In addition, it also supports PROC SQL views and MDDBs (Multi-dimensional Data Base) for read-only access.
Update access is not supported. (Input and Output access is supported.)
There is no WHERE expression optimization with an index.
CEDA is limited to directory-based platforms (bound libraries on OS/390 and CMS files are not part of CEDA).

If your application can operate within these restrictions, then CEDA is a simpler cross-platform data access strategy than the Version 6 data transfer services and remote library services. If your needs go beyond these restrictions, then data transfer services and remote library services are still available to provide access to remote data across all platforms and releases of SAS software.

A user who does not have OS/390 cannot use CEDA to reference an OS/390 bound library. However, Version 7 or Version 8 SAS software supports unbound (or directory-based) libraries processed with the OS/390 UNIX Support Services. As an OS/390 user, you can use the IBM NFS Client to get to UNIX files from an OS/390 SAS session. As a UNIX user, you can use the IBM NFS Server to reference OS/390 UNIX Support Services files from a UNIX platform. CEDA performs the necessary translations so that these cross platform references seem like local references.

The number of observations may limit the usability of a data set transferred between two heterogeneous hosts using the CEDA facility. The reason this is a limitation is because the number of observations in a data set is stored in a long numeric type, and the number of bytes that are used to represent a long numeric type varies between hosts. The minimum is 4 bytes. Therefore, if you have data sets that contain more than 2**31-1 observations, you will not be able to access this data set on a host that stores a long numeric type in 4 bytes. This maximum observations limit may also be influenced by having an index or integrity constraints that use an index.

The failure that occurs due to the maximum observation limit is either you cannot open the data set, or you cannot add additional observations. There are four conditions where this error may occur:

if a 4-byte long host opens an 8-byte long data set that already contains greater than 2**31-1 observations
if an 8-byte long host opens a 4-byte long data set that already contains greater than 2**31-1 observations
if a 4-byte long host attempts to add observation 2**31-1 in an 8-byte long data set
if an 8-byte long host attempts to add observation 2**31-1 in an 4-byte long data set.

Chapter Contents
Previous
Next
Top of Page