Browse Prior Art Database

Rolling Checkpoint in a Transaction Processing System

IP.com Disclosure Number: IPCOM000040510D
Original Publication Date: 1987-Nov-01
Included in the Prior Art Database: 2005-Feb-02
Document File: 3 page(s) / 73K

Publishing Venue

IBM

Related People

Narang, IS: AUTHOR [+2]

Abstract

A method is described whereby a resource manager (RM) in a transaction processing system takes a checkpoint of its resources as follows: (a) progressively takes a checkpoint of a subset of resources; (b) records this partial checkpoint data in a checkpoint dataset; (c) repeats steps (a) and (b) until all resources are checkpointed; and (d) repeats step (c) on the dataset other than the one on which the checkpoint just completed. There are two datasets which are used alternately to take checkpoints. The RM records the modification of its resource in a journal. Each journal record has a sequence number. RM maintains the highest running sequence number across all resources. At the time RM scans the resource for its checkpoint data, all modifications to the resource up to the current highest sequence number are included.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 3

Rolling Checkpoint in a Transaction Processing System

A method is described whereby a resource manager (RM) in a transaction processing system takes a checkpoint of its resources as follows: (a) progressively takes a checkpoint of a subset of resources; (b) records this partial checkpoint data in a checkpoint dataset; (c) repeats steps (a) and (b) until all resources are checkpointed; and (d) repeats step (c) on the dataset other than the one on which the checkpoint just completed. There are two datasets which are used alternately to take checkpoints. The RM records the modification of its resource in a journal. Each journal record has a sequence number. RM maintains the highest running sequence number across all resources. At the time RM scans the resource for its checkpoint data, all modifications to the resource up to the current highest sequence number are included. This sequence number is referred to as the Cutoff Sequence Number (CSN) and is noted with the checkpoint data for the resource. The checkpoint record for a resource consists of its resource-id, CSN, length, a record-continuation indicator and the checkpoint data. The RM checkpoints the resources in a specific order. Two datasets are used alternately for taking the checkpoints. The one on which the checkpoint is being taken currently is called the current checkpoint dataset. A checkpoint is assigned a monotonically increasing checkpoint id. The checkpoint id is recorded in each block (of the dataset) as the checkpoint is taken progressively. All the resources whose records fit completely in the block are considered checkpointed when that block is written. A resource whose record fits in more than one block is considered checkpointed when the block containing its last piece (indicated by the record-continuation indicator) is written. The last block in the dataset is marked with the last-block indicator to indicate that a complete checkpoint exists on the dataset. The figure illustrates a...