Problem Note 43966: Results of query against a very large SAS® Scalable Performance Data Server® table might be incorrect
The following example shows code that might return incorrect results in
SAS® Scalable Performance Data
Server® when executed against
a very large table:
proc sql;
create table target as
select sum(a), sum(b), sum(c)
from spdslib.large_table
group by a, b , c
;
quit;
The SAS Scalable Performance Data Server parallel GROUP BY threads
hold a partial GROUP BY in memory, based on the GRPBYROWCACHE server
parameter.
Each thread is given GRPBYROWCACHE/MAXWHTHREADS memory.
When the memory is full, each thread flushes the partial GROUP BY to disk in a per-thread spill file.
The final merge of the parallel GROUP BY reads the per-thread partial results
back in.
Incorrect results might occur on the final merge for a very large table.
Select the Hot Fix tab in this Note to access the hot fix for this problem.
Operating System and Release Information
SPDS | Scalable Performance Data Server | Microsoft Windows Server 2003 Datacenter 64-bit Edition | 4.52 | 4.52 HF5 | | |
Microsoft Windows Server 2003 Enterprise 64-bit Edition | 4.52 | 4.52 HF5 | | |
Microsoft Windows Server 2003 Datacenter Edition | 4.52 | 4.52 HF5 | | |
Microsoft Windows Server 2003 Enterprise Edition | 4.52 | 4.52 HF5 | | |
Microsoft Windows Server 2003 Standard Edition | 4.52 | 4.52 HF5 | | |
Microsoft Windows XP Professional | 4.52 | 4.52 HF5 | | |
64-bit Enabled AIX | 4.52 | 4.52 HF5 | | |
64-bit Enabled HP-UX | 4.52 | 4.52 HF5 | | |
64-bit Enabled Solaris | 4.52 | 4.52 HF5 | | |
Linux for x64 | 4.52 | 4.52 HF5 | | |
Solaris for x64 | 4.52 | 4.52 HF5 | | |
*
For software releases that are not yet generally available, the Fixed
Release is the software release in which the problem is planned to be
fixed.
A hot fix is available for this problem.
Type: | Problem Note |
Priority: | alert |
Topic: | System Administration ==> Servers ==> SPDS
|
Date Modified: | 2011-08-15 18:41:31 |
Date Created: | 2011-08-11 08:55:46 |