JPS581242A - Storage controlling system for tree structure information - Google Patents

Abstract

PURPOSE:To remarkably reduce the processing time, by possibly locating node information corresponding to a node located to an upper hierarchy of the tree structure into a resident region, in a tree structure retrieval processing device. CONSTITUTION:In a tree structure retrieval processing system in which a tree structure storage information storing section is constituted in page units respectively stored in a resident area 4 and a nonresident area 5 on a main storage device, the information corresponding to the node of upper hierarchy of the tree structure is stored to the resident area 4 in the order of higher accessing frequency, and a plural-hierarchy information is assigned to a page unit transferred and stored to the nonresident area 5 so that the node located at the lower hierarchy takes a pair to one node in the page unit, allowing to execute the roll-in/ roll-out processing between the nonresident area 5 and an external storage area. Thus, the frequency of generation of page fault can remarkably be reduced and the number of times of roll-in/roll-out with the external storage device can be decreased.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides a tree-structured information storage control system, particularly a tree-structured KmH
In a memory structure that stores information stored in a main memory, a page-by-page management structure is used so that the information resident in the main memory corresponds to a higher layer of the main structure, and undesired page faults are avoided. This invention relates to a log structure information storage control method that reduces the frequency of occurrence.

Conventionally, correspondence relationships expanded into a tree structure have been stored in an information storage unit, for example in the case of an English dictionary, where output information is falsely associated with a plurality of columns of key information. The information storage unit is accessed based on the nine search code strings input during the search, and output information is extracted while comparing the key information with the codes on the search code string.

Such tree-structured information is composed of a set of key information given corresponding to a node and a pointer indicating which lower node to reach from that node, for example, in units of one word. Information in units of one word is collected in units of one page and incorporated into a data processing system having a virtual storage method.

When summarizing the information in the upper volume structure on a page-by-page basis,
For example, in the case of a tree structure in which four branches occur from one node, for example, information about one node, four nodes located under it, and 16 nodes located further under it, that is, the total A method is adopted in which information regarding 21 nodes is configured as a 1-1 page.

In this method, when one node is transferred to the main memory, there is a high possibility that the nodes in the lower hierarchy will also be transferred to the main memory, and after accessing the second layer, the +1) This can be considered to be based on the idea that the frequency of undesired page/7-orts occurring when attempting to access a hierarchy will be reduced.

However, when the page units divided as described above are distributed on the main storage device and the external storage device, 1llp
If there are four resident areas on the K main memory and you try to make the above page units resident in the main permanent area, a problem will occur as described later with reference to Figure 1, and a page fault will occur. It happens that the number of times increases.

The present invention aims to solve this problem, and stores information corresponding to nodes higher up in the tree structure in a permanent area or an area that is less likely to be rolled out. In this way, the output information is stored in this structure in which key information and one or more pointers are set corresponding to the synode, and the output information is stored in multiple key codes. Tozu〈A tree structure search that extracts the above output information while comparing each code of the code string with the key information set corresponding to the above node is also performed in the logical device. The tree-structured information storage section includes a tree-structured storage information storage section that stores pointers, and the tree-structured information storage section is a tree configured by page units stored in a resident area on the main storage device and a non-resident area on the main storage device, respectively. In the structure search processing system, the information corresponding to the upper layer nodes of the main structure is also stored in the resident area in descending order of access frequency, and the information is also transferred and stored in the non-resident area. Regarding page units, information of multiple hierarchies is allocated so that nodes located in lower hierarchies are paired with one node within the page unit, and roll-in/- - The feature is that the loop-out processing is extended in depth. This will be explained below with reference to the drawings.

FIG. 1 shows an example of a conventional storage control system.

Figure 2 is an embodiment of the present invention, Figure 3 (4) (B) (C)
(b) shows nine explanatory views related to the present invention.

In FIG. 1, 1-1.1-2. . . . represent a page unit, 2 a node, 3 a path, 4 a resident area, and 5 a non-resident area and an external storage area.

The illustrated node 2 represents the nine node information in which the key information and pointer corresponding to the node are grouped together.

Now consider a tree structure with four branches from one node 2, and consider that 12 bytes of key information and 3×4+ bytes can be used as pointers to the four lower nodes, including each node information. Assuming that 9811 bytes of information can be accommodated per page, as shown in Figure 1, a maximum of 21 pieces of node information can be accommodated in one page, so the number of pieces of node information belonging to the hierarchy is one. (p+1)th
One page is composed of four pieces of node information belonging to a layer and 16 pieces of node information belonging to the (p+2>8th layer) (where 1m8m-2:s is a natural number).

Assuming that the tree structure of the book is currently composed of 8 levels, and that the book can accommodate 16 pages in the resident area 4 of the main memory, the tree structure is counted from the top level as shown in Figure 1. Page 1-1, which summarizes the content up to the third level, pages 1-2 to 1-16, which summarize the content of three levels including the fourth level, are accommodated in the resident area, and the following pages 1-17 ．．・
..., 1-65.

......, 1-69. ......, 1-416
1 is arranged in a non-resident area on the main storage device and an external storage area on the external storage device.

Figure 1 shows this state. Pages in the non-resident area can be freely rolled in and out of the external storage area, as is well known when a virtual storage system is adopted.・Outsale.

The structure shown in FIG. 1 As can be seen, in pages 1-2 to 1-IN, node information corresponding to #- for nodes in the 5th layer and nodes in the 6th layer is resident in the resident area 4. However, on pages 1-1 to 1-65, the node information corresponding to the nodes in the fourth layer is in a non-resident state.

Now, as shown in Figure 1, consider a tree in which the number of nodes on the path from the root of the tree to each end node is equal across all paths, and the frequency of searching on each path is approximately equal across all buses. Assuming that the existence of node information corresponding to a node in a higher layer of the tree structure in the main memory is the way to reduce the number of page fault occurrences described above, in the case of the configuration shown in Figure 1. (assuming that the end nodes of the tree structure are searched on average), there is a form in which on average 41-1515 2 X -41x-selfL78(times) 343 page faults occur per search. becomes.

FIG. 2 shows an embodiment of the present invention, and the reference numeral 1-1 in the figure shows an embodiment of the present invention.
, 1-2, . . . , 2.3.4 correspond to FIG.

In the case shown in Figure 2, the page resident in resident area 4 is configured based on the idea that node information corresponding to nodes located at higher levels of the tree structure is placed in resident area 4 as much as possible. I'm so busy.

In this way, it becomes possible to adopt a configuration in which node information corresponding to nodes up to the fifth layer resides in the resident area 4, as shown in FIG. 2 (0). Then, the node information corresponding to the nodes of the 6th layer and below is stored in one page, and the node information corresponding to a total of 21 nodes up to the 8th layer is stored between the non-resident area and the external storage area. Roll in/roll out as needed.

By placing up to the fifth layer in the resident area 4 as shown in FIG. 2, the average number of page faults per search is one at the worst, which is extremely effective.

It has been shown that in the case of the configuration shown in FIG. 2, the frequency of occurrence of page 7 oorts is reduced. However, in all cases, the configuration shown in FIG. 2 does not necessarily reduce the frequency of occurrence of errors and faults compared to the configuration shown in FIG. That is, the structure of the current tree is modeled as shown in the third drawing, so that the shaded area can be stored in the resident area, and the hierarchy is set in dotted page units. At this time, Figure 3 (B)
In the case shown and in the case shown in Figure 3 (C1),
B) The case shown is more preferable. This is because in the former case, page faults may occur twice or three times, but in the latter case, page faults occur tia times. However, if the configuration shown in Figure 3 (CB) can be made into the configuration shown in Figure 3 (B), the configuration shown in Figure 3 (To) is preferable.

As explained above, according to the present invention, when extracting output information, the frequency of page faults occurring is significantly reduced, and the number of times of roll-in/roll-out to/from an external storage device is significantly reduced. processing time is significantly reduced.

[Brief explanation of drawings]

FIG. 1 shows an example of a conventional storage control system, FIG. 2 shows an embodiment of the present invention, and FIG. In the figure, 1 is page unit, 2#i-node, 4 is resident area, 5
indicates non-resident area and external storage area. Unexamined Japanese Patent Publication No. 58-1242 (4) Illustration of two twirling spears

Claims (1)

[Claims] Output information is not stored in a tree structure in which key information and one or more pointers are set corresponding to the nodes, and each code of a code string based on multiple key codes is stored in a tree structure in which key information and one or more pointers are set corresponding to the nodes. The present structured search processing system for extracting the output information while comparing with the set key information is provided with a tree structure storage information storage unit storing the key information and the pointer based on the tree structure, In a tree structure search processing system in which a tree structure storage information storage section is configured of pages stored in a permanent area on the main storage device and a non-resident area on the main storage device, the above page is stored in the permanent area on the main storage device. Is the information corresponding to the nodes in the upper layer of the tree structure frequently accessed? In addition, regarding the page unit transferred and stored in the non-resident area, information of multiple hierarchies is stored so that one node is paired with a node located in a lower hierarchy within the page unit. A tree-structured information storage control method, characterized in that roll-in/roll-out processing is executed between the non-resident area and the external storage area.