Constructing a transaction serialization order based on parallel or distributed database log files

• US 5,878,414 A
• Filed: 06/06/1997
• Issued: 03/02/1999
• Est. Priority Date: 06/06/1997
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Hsiao, Hui-I, Lo, Ming-Ling
• Primary Examiner(s): Black, Thomas G.
• Assistant Examiner(s): MIZRAHI, DIANE D
• Application Number:
  US08/871,001
• Time in Patent Office: 634 Days
• Field of Search: 707/8, 707/1, 707/9, 707/10, 707/3, 395/183.01, 395/726, 600/316, 370/259
• US Class Current: 707/648
• CPC Class Codes: G06F 16/2358 Change logging, detection, and notification replication G06F16/27
• CPC Class Codes: Y10S 707/959 Network
• CPC Class Codes: Y10S 707/966 Distributed
• CPC Class Codes: Y10S 707/99931 Database or file accessing
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99938 Concurrency, e.g. lock management in shared database
• CPC Class Codes: Y10S 707/99939 Privileged access

Abstract

A method to construct a transaction serialization order based on parallel or distributed database log files that connects the log files into a network and merges the network into a sequence. Only vote symbols and commit symbols are considered in the construction and a protocol of a transaction'"'"'s vote appearing before a transaction'"'"'s commit is enforced. The log files are connected by making links between commit symbols of the same transactions in different log files. The method identifies ambiguities in the orders of the individual log files, called commit symbol cycles, and resolves them by breaking an unsupported segment of the cycle and sliding the segment'"'"'s forwardmost commit symbol backward to a position immediately prior to the first commit symbol on that segment. By doing so, the serialization implication of an individual log file may be changed, but the overall serialization implication of the database remains intact. Using an iterative process, this method can order the log files in one pass-through.

Claims

1. A method for constructing a list having a transaction serialization order using at least two local log files in a parallel or distributed database system comprising the steps of:

creating a network of log files by identifying each commit symbol of each transaction in a first local log file with a concomitant commit symbol of said transaction in other local log files having a commit symbol associated with said transaction to thereby designate merge points;

traversing said network; and

merging said network into a single sequence, using the designated merge points.

2. The method according to claim 1 wherein said traversing comprises the steps of:

(a) examining a local log file from the beginning until said merge point is encountered;

(b) recording in a wait list of said local log file the names of local log files associated with said merge point;

(c) if all local log files have not been completely examined, finding a local log file with no local log file names in its wait list and continuing with step (a);

(d) if all local log files have local log file names in their wait lists, resolving a commit symbol cycle and continuing with step (c);

(e) continuing until all local log files have been examined.

3. The method according to claim 2 wherein said merging further comprises the steps of:

determining, after the names of local log files connected to said merge point have been recorded in said wait list of said local log file, whether said local log file names mutually appear in said wait lists of said local log files for said merge point;

merging local log files up to said merge point in each local log file in which said local log file names mutually appear in other local log files'"'"' wait lists; and

deleting from each local log file'"'"'s wait list the names of mutual local log files that have been merged.

4. A method for constructing a list having a transaction serialization order using a database log file comprising the steps of:

requiring a vote symbol of a transaction in said log file to occur at a time earlier than a commit symbol of said transaction in said log file;

examining the placement of a commit symbol of a first transaction and a vote symbol of a second transaction and a commit symbol of a second transaction;

wherein said first transaction is ordered before said second transaction if said commit symbol of said first transaction occurs at a time earlier than said vote symbol of said second transaction.

5. A method for adjusting the position of a type of operation symbol of a first transaction of a database log file relative to the position of said type of operation symbol of a second transaction of said database log file wherein said type of operation symbol of said first transaction occurs later in time than said type of operation symbol of said second transaction comprising the step of moving said position of said type of operation symbol of said first transaction to a position earlier than that of said type of operation symbol of said second transaction.

6. The method according to claim 5 wherein said position adjustment neither creates nor destroys a serialization implication of said database log file.

7. The method according to claim 5 wherein said type of operation symbol is a commit symbol.

8. A method for identifying a commit symbol cycle in database log files comprising the steps of:

identifying a first local log file and at least a second local log file;

identifying a commit symbol of a first transaction and a commit symbol of a second transaction on said first local log file and denominating a connection between said two commit symbols as a first segment;

identifying a commit symbol of said first transaction and a commit symbol of said second transaction on said second local log file and denominating a connection between said two commit symbols as a second segment;

connecting via a first temporary connection said commit symbol of said first transaction on said first local log file to said commit symbol of said first transaction on said second local log file;

connecting via a second temporary connection said commit symbol of said second transaction on said first local log file to said commit symbol of said second transaction on said second local log file;

wherein traversing in a forward direction on said first segment from the beginning of said first segment to the end of said first segment and then traversing via a said second temporary connection from the end of said first segment to the end of said second segment and then traversing in a forward direction on said second segment leads to said commit symbol of said first transaction on said second local log file.

9. A method for resolving a commit symbol cycle in database log files comprising the steps of:

identifying a commit symbol cycle and a first segment thereof;

determining an unsupported segment by examining if said first segment is unsupported and, if said first segment is not unsupported, continuing to each next segment of said commit symbol cycle until an unsupported segment is found;

breaking said unsupported segment between a commit symbol of a first transaction and a commit symbol of a second transaction wherein said commit symbol of said second transaction is in a position earlier in said unsupported segment than said commit symbol of said first transaction; and

sliding said commit symbol of said first transaction backward to a position immediately prior to said commit symbol of said second transaction.

10. The method according to claim 9 wherein a segment is determined by a further method comprising the step of denominating a connection between a commit symbol of a first transaction and a commit symbol of a second transaction on a log file.

11. The method according to claim 10 wherein said segment is determined to be unsupported by a further method comprising the steps of:

determining if said segment is directly supported; and

if said segment is not directly supported, determining whether it is indirectly supported;

wherein if said segment is neither directly supported nor indirectly supported, it is unsupported.

12. The method according to claim 11 wherein said segment is determined to be directly supported by a further method comprising the step of traversing in a forward direction along said log file beginning at said commit symbol of said first transaction and ending at said commit symbol of said second transaction, wherein if a vote symbol of said second transaction is encountered before said commit symbol of said second transaction, said segment is directly supported.

13. The method according to claim 11 wherein said segment on a first log file is determined to be indirectly supported by a further method comprising the step of identifying on a second log file a commit symbol of said first transaction, wherein said commit symbol of said first transaction on said second log file is positioned before a vote symbol of said second transaction on said second log file, and wherein said vote symbol of said second transaction on said second log file is positioned before a commit symbol of said second transaction on said second log file.