8. Configuring and Managing Vector : X100 Configuration Parameters : [memory] Settings
 
Share this page                  
[memory] Settings
Memory settings in vectorwise.conf affect the memory used for query processing (not for data caching in the buffer pool).
max_memory_size
Specifies the amount (in bytes) of the total memory used for query execution.
Notes:
Max_memory_size applies to all queries executing across the system, but a single query can consume almost all available memory.
Some queries will fail if they consume too much memory; increasing available memory can help.
If max_memory_size is not specified or is not larger than zero, a default setting of 50% of the physical system memory is used.
Memory size defined with this option does not include bufferpool_size.
For more information, see Memory Settings.
Limits: 32 MB to 256 TB
Default: 0 (use 50% of the physical system memory)
max_overalloc
Specifies the maximum amount of memory that will be overallocated (reserved as virtual memory) for zero-cost growing of certain memory areas. The setting needs to be changed only in rare cases.
A value of 0 means no overallocation. A value of ‑1 means overallocate with the amount of total system memory ("unlimited").
Note:  Overallocation should be disabled on operating systems where: /proc/sys/vm/overcommit_memory is not 1, or "ulimit -v" is not unlimited.
Default: 2G or set to 0 upon installation if the system does not allow overallocation.
huge_tlb_lim
Enables the huge pages feature of the CPU only for queries using the number of cores less than or equal to the specified value. Set this option to 1 if parallel queries do not benefit from the huge pages feature. This may be experienced on some operating systems, such as Red Hat Enterprise Linux 6, which has an automatic anonymous TLB feature.
Special value 0 means there is no limit.
Default: 0 (no limit)
use_huge_tlb
Specifies whether to use the huge pages feature (see Using Large Pages) of the CPU. Valid values are true and false.
Default: false
use_numa
Specifies whether to use NUMA optimization (see Using NUMA Optimization). Valid values are true and false.
Default: true