Efficient performance of insert and point query operations in a column store

A computer program product for moving hierarchical data objects across two or more kinds of memories, the computer program product comprising a computer readable storage medium having program code embodied therewith, the program code executable by a processor to:

• logically organize, by the processor, data objects in a first hierarchy, wherein a portion of the data objects in the first hierarchy logically comprise groupings of other data objects, each logical grouping containing a certain number of contiguous logical records, identified individually by tuple identifiers as tuple sequence numbers (TSNs);
• physically organize, by the processor, the data objects across two or more types of memory in a second hierarchy, wherein another portion of the data objects in the second hierarchy physically comprise groupings of other data objects;
• dynamically move, by the processor, groupings of the data objects in the second hierarchy across the two or more types of memory;
• track levels, by the processor, of access of groups of the data objects using an across-column data structure, organized at a particular number of rows, that maps ranges of tuple sequence numbers of the data objects in the first hierarchy onto metadata information including combined access frequencies of the data objects, and current number of accessors to the data objects, in each grouping of the data objects, wherein the across-column data structure comprises, for each column of data:
• an array pair of a starting tuple sequence number (startTSN) and pointer (pagePTR) to a memory page (startTSN, pagePTR), the startTSN is a TSN of the starting TSN in each memory page for a particular column of data, a pagePTR is a union of pointers across the first hierarchy and the second hierarchy, the across-column data structure stores a direct mapping from an identifier for a particular row to a virtual memory address of all data pages straddling the particular row in any column are stored in the across-column data structure, and coarse reference counting, to record all active reference to rows of the across-column data structure, is performed outside of any database buffer pool in the across-column data structure while maintaining separate reference counts for each column;
• receiving a query with a list of columns for the table:
• incrementing the access count for each of the columns and a global reader counter in the across-column data structure;
• scanning (startTSN, pagePTR) for each column of the list of columns;
• identifying a targeted page based on a determination that the targeted page is present in the first hierarchy;
• and
• reading values from the target page and returning to the query.

A system comprising:

• a processor;
• two or more storage units coupled to the processor, wherein the two or more storage units comprise different memory types;
• an object hierarchy processor that:
• logically organizes data objects in a first hierarchy, wherein a portion of the data objects in the first hierarchy logically comprise groupings of other data objects, each logical grouping containing a certain number of contiguous logical records, identified individually by tuple identifiers as tuple sequence numbers (TSNs),
• physically organizes the data objects across the two or more storage units in a second hierarchy, wherein another portion of the data objects in the second hierarchy physically comprise groupings of other data objects, and
• dynamically moves groupings of the data objects in the second hierarchy across the two or more storage units;
• an access tracking processor that tracks levels of access of groups of the data objects using an across-column data structure, organized at a particular number of rows, that maps ranges of tuple sequence numbers of the data objects in the first hierarchy onto metadata information including combined access frequencies of the data objects, and current number of accessors to the data objects, in each grouping of the data objects, wherein the across-column data structure comprises, for each column of data:
• an array pair of a starting tuple sequence number (startTSN) and pointer (pagePTR) to a memory page, the startTSN is a TSN of the starting TSN in each memory page for a particular column of data, a pagePTR is a union of pointers across the first hierarchy and the second hierarchy, and the across-column data structure stores a direct mapping from an identifier for a particular row to a virtual memory address of all data pages straddling the particular row in any column are stored in the across-column data structure, and coarse reference counting, to record all active reference to rows of the across-column data structure, is performed outside of any database buffer pool in the across-column data structure while maintaining separate reference counts for each column;
• receiving a query with a list of columns for the table;
• incrementing the access count for each of the columns and a global reader counter in the across-column data structure;
• scanning (startTSN, pagePTR) for each column of the list of columns;
• identifying a targeted page based on a determination that the targeted page is present in the first hierarchy;
• and
• reading values from the target page and returning to the query.