Technique for stabilizing data in a non-log based information storage and retrieval system

• US 20020103819A1
• Filed: 12/12/2000
• Published: 08/01/2002
• Est. Priority Date: 12/12/2000
• Status: Abandoned Application

• Current Assignee: Fresher Information Corporation
• Original Assignee: Fresher Information Corporation
• Inventors: Cabannes, Didier, Duvillier, Edouard
• Application Number:
  US09/735,819
• Publication Number:
  US 20020103819A1
• Time in Patent Office: Days
• US Class Current: 707/206
• CPC Class Codes: G06F 16/22 Indexing; Data structures therefor; Storage structures
• CPC Class Codes: G06F 16/2329 using versioning
• CPC Class Codes: G06F 16/2358 Change logging, detection, and notification replication G06F16/27
• CPC Class Codes: G06F 16/24 Querying

Abstract

A technique for stabilizing and collecting data in an information storage and retrieval system, referred to as checkpointing, is described. Checkpointing is used to increase the speed of the database during a recovery by only scanning data that the information storage and retrieval system knows is unstable, instead of scanning all the data in the database. Data that is deemed collectable, such as old data or obsolete data, is identified in a non-persistent memory space, such as a cache memory. A data page contained in an initial or first buffer is stored, also in the form of a data page, to a persistent memory type, such as a hard drive or virtual memory. Next, non-collectable data, or data that is to be maintained, in the initial or first buffer is identified. This data is stored in a second buffer. It is then determined whether the non-collectable data is referenced in an object table in the information storage and retrieval system. A first checkpoint flag field in an allocation map in the non-persistent memory area is set. Once the checkpoint flag field is set, the second buffer is flushed to the non-persistent memory type.

Claims

1. A method of collecting data in an information storage and retrieval system comprising:

identifying collectable data in a first memory type;

storing a data page in a first buffer in a second memory type;

identifying non-collectable data in the first buffer and storing the non-collectable data in a second buffer;

determining whether the non-collectable data is referenced in an object table;

setting a first checkpoint flag field in an allocation map in the first memory type; and

flushing the second buffer to the first memory type.

2. A method as recited in claim 1 further including setting a second checkpoint flag field in a header for the second buffer.

3. A method as recited in claim 1 further including obtaining at least one first memory type address for the non-collectable data in the flushed second buffer and storing the first memory type address in the header of the second buffer.

4. A method as recited in claim 1 wherein the at least one first memory type address is obtained at an optimal speed for hardware being used by the information storage and retrieval system.

6. A method as recited in claim 1 further comprising:

selecting a data page in the first buffer;

determining if a first checkpoint flag field corresponding to the selected data page is set in the allocation map;

if the first checkpoint flag field is not set, setting a free flag field in the allocation map; and

if the first checkpoint flag field is set, setting a to-be-released flag field in the allocation map.

7. A method as recited in claim 1 wherein the allocation map has a corresponding data page.

8. An information storage and retrieval system capable of intrinsic versioning of data comprising:

a disk header having an object table root address and an allocation map area address;

an allocation map area having at least one allocation map having a checkpoint flag field;

a stable data segment having a current persistent object table, a saved object table, and stable data; and

an unstable data segment containing unstable data.

9. An information storage and retrieval system as recited in claim 8 wherein an allocation map has a free flag field, a to-be-released flag field, and a page identifier field.

10. A method of stabilizing a database comprising:

flushing an object table from a first memory type to a second memory type;

migrating a checkpoint flag from a first allocation map to a second allocation map in the first memory type;

moving the second allocation map to the second memory type; and

updating a header of the second memory type to indicate a location of the object table and the second allocation map.

11. A method as recited in claim 10 further comprising:

scanning the second allocation map to identify a data page having a corresponding to-be-released flag that has been set in the second allocation map; and

resetting the corresponding to-be-released flag and setting a corresponding free flag in the second allocation map for the identified data page.

12. A method of stabilizing a non-log based database, the method comprising:

determining which data has not been stabilized by examining a checkpoint flag, wherein the data is in the form of an object version having one of a transaction identifier and a version identifier;

determining if the object version is mapped to an object table; and

if the object version is mapped to the object table, setting the checkpoint flag for the object version, thereby designating the object version as stable data and ignorable data when rebuilding the object table after a restart of the database.

13. A computer program product of collecting data in an information storage and retrieval system comprising:

a computer usable medium having computer readable code embodied therein, the computer readable code comprising;

computer code for identifying collectable data in a first memory type;

computer code for storing a data page in a first buffer in a second memory type;

computer code for identifying non-collectable data in the first buffer and storing the non-collectable data in a second buffer;

computer code for determining whether the non-collectable data is referenced in an object table;

computer code for setting a first checkpoint flag field in an allocation map in the first memory type; and

computer code for flushing the second buffer to the first memory type.

14. A computer program product of stabilizing a non-log based database, the computer program product comprising:

a computer usable medium having computer readable code embodied therein, the computer readable code comprising;

computer code for determining which data has not been stabilized by examining a checkpoint flag, wherein the data is in the form of an object version having one of a transaction identifier and a version identifier;

computer code for determining if the object version is mapped to an object table; and

computer code for setting the checkpoint flag for the object version if the object version is mapped to the object table, thereby designating the object version as stable data and ignorable data when rebuilding the object table after a restart of the database.

15. A computer program product for stabilizing data in a database, the computer program product comprising:

a computer usable medium having computer readable code embodied therein, the computer readable code comprising;

computer code for flushing an object table from a first memory type to a second memory type;

computer code for migrating a checkpoint flag from a first allocation map to a second allocation map in the first memory type;

computer code for moving the second allocation map to the second memory type; and

computer code for updating a header of the second memory type to indicate a location of the object table and the second allocation map.

16. A system for collecting data in an information storage and retrieval system comprising:

means for identifying collectable data in a first memory type;

means for storing a data page in a first buffer in a second memory type;

means for identifying non-collectable data in the first buffer and storing the non-collectable data in a second buffer;

means for determining whether the non-collectable data is referenced in an object table;

means for setting a first checkpoint flag field in an allocation map in the first memory type; and

means for flushing the second buffer to the first memory type.