Computer program product to enable multiple computer systems to share single ...

COMPUTER PROGRAM PRODUCT TO
ENABLE MULTIPLE COMPUTER SYSTEMS
TO SHARE SINGLE SEQUENTIAL LOG

This application is a division of U.S. application Ser. No. 5 08/394,582 filed Feb. 27 1995, (status: pending).

TECHNICAL FIELD

The present invention relates generally to sequential logs, and more particularly to a system and method for enabling 10 multiple computer systems to share a sequential log.

BACKGROUND ART

As is well known, a sequential log (also herein called "sequential log file," "sequential file log," or simply "log") 15 is a database wherein each new data record is stored in the next available (i.e., free) physical memory location of the log. In other words, in a sequential log, new data records are stored in sequential physical memory locations of the log.

For example, suppose that a computer system generates 20 the following data records (in the indicated order): record-A, record-B, and record-C. If the computer system writes these data records to a sequential log in the order in which the data records are created, then record-A would be stored in physical location N of the log, record-B would be stored in 25 physical location N+l of the log, and record-C would be stored in physical location N+2 of the log (where physical location N is the first available memory location in the log).

Consider FIG. 1, which illustrates an example computer system 102 and an example sequential log 106. The sequential log 106 includes a plurality of physical memory blocks 108A-108E, where each memory block 108 is of a size to store three data records. For example, memory block 108A stores data records A, B, and C. (The letter designations of data records indicate their sequence of creation. Thus, data record A was created first, data record B was created second, data record C was created third, etc.)

The computer system 102 includes a buffer 104 (implemented using random access memory, RAM, or a 4Q combination of registers, for example). The buffer 104 is of a size to store three data records.

The computer system 102 stores data records in the buffer 104 as such data records are created. In the example of FIG. 1, the buffer 104 has stored therein data records G and H. 45 When data record I is created, it shall be stored in the remaining slot of the buffer 104, at which time the buffer 104 shall be full.

When the buffer 104 becomes full, its contents are stored in the next available (free) block in the sequential log 106. 50 In the example shown in FIG. 1, the next available block in the sequential log 106 is block 108C. Thus, when the buffer 104 becomes full, its contents are stored in physical memory block 108C, as indicated by line 110.

As should be clear from the above discussion, data 55 records are stored in a sequential log in the order in which such data records are created. In the example of FIG. 1, data record A was the first to be created, and is thus stored in the first memory location of the log 106. Data record B was created immediately after data record A, and is thus stored 60 in the next memory location of the log 106.

This characteristic of sequential logs makes them well suited for use with applications where data records must be stored in the order in which they are created. Such applications include financial institutions (such as banks and loan 65 companies), insurance companies, package delivery services, food delivery services, etc.

30

35

Traditionally, a sequential log is dedicated to a single computer system, for the exclusive use by that computer system. This is the scenario presented in FIG. 1. With the advent of distributed and parallel processing, however, it has become advantageous to allow multiple computer systems (connected over a computer network, for example) to have access to a single sequential log.

Consider, for example, a banking scenario where a bank has a first branch and a second branch, the first and second branches being located in different cities. Suppose that a customer makes a withdraw at the first branch, and then later in the day makes a deposit at the second branch, and then still later in the day makes a withdraw at the first branch. It is preferable that the two computer systems (corresponding to the first and second branches) share the same sequential log. Otherwise, the first branch will have no knowledge of the deposit made at the second branch, such that the second withdraw at the first branch may be denied due to insufficient funds.

There are problems associated with allowing multiple computer systems to have access to a single sequential log. This is illustrated in FIG. 2, where a first computer system 202 and a second computer system 206 share a sequential log 210. The first system 202 temporarily stores data records in its internal buffer 204. Similarly, the second system 206 temporarily stores data records in its internal buffer 208. Note that the first system 202 has generated and stored data record A in its buffer 204. The second system 206 has then generated and stored data records B and C in its buffer 208. The first system 202 has then generated and stored data record D in its buffer 204.

The sequential log 210 stores data records it receives in sequential physical memory blocks, irrespective of the source of such data records. Suppose that the second system 206 generates and stores data record E in its buffer 208, and then sends the contents of its buffer 208 to the sequential log 210. Then, the first system 202 generates and stores data record F in its buffer 204, and sends the contents of its buffer 204 to the sequential log 210. In this scenario, the contents of buffer 208 would be stored in block 212A (as indicated by line 216), and the contents of buffer 204 would be stored in block 212B (as indicated by line 214).

According to this scenario, the contents of the sequential log 210 would be as follows (starting from the top of the log 210): B, C, E, A, D, F. Note that the data records are not stored in the sequential log 210 in the order in which they were created. Thus, the use of a sequential log with multiple computer systems results in improper behavior.

FIG. 3 illustrates one conventional approach for solving this problem, where each computer system is provided with its own sequential log. In the example of FIG. 3, a first computer system 302 is provided with a first sequential log 310, and a second computer system 306 is provided with a second sequential log 314. When buffer 304 in the first computer system 302 becomes full, its contents are stored in the next available block in the first log 310. Similarly, when buffer 308 in the second computer system 306 becomes full, its contents are stored in the next available block in the second log 314. At predetermined time intervals (such as the end of the day), the contents of the first and second logs 310, 314 are merged and sorted.

The approach shown in FIG. 3 is flawed for a number of reasons. First, this approach is very expensive, since each computer system must be provided with its own sequential log. Second, this approach is very complex, because the operation of all of the logs must be coordinated (so as to read out and merge the data contained in such logs).

5

only if record M arrives at the shared memory 404 before record N. (Such race condition problems can be eliminated by adding well known handshaking/protocol mechanisms.)

The inventors have developed another approach for enabling multiple computer systems to access a single 5 sequential log. This approach, which represents the preferred embodiment of the present invention, shall now be described with reference to FIG. 5.

According to the preferred embodiment of the present invention, a plurality of computer systems (such as a first 10 computer system 502, a second computer system 506, . . . , and an Nth computer system 510) have access to a common database system 514. The computer systems 502, 506, 510 also have access to a coordinate clock 522 (described below). Each of the computer systems 502, 506, 510 has an 15 internal buffer 504, 508, 512, respectively (implemented using random access memory, RAM, or a series of registers, for example). The computer systems 502, 506, 510 may be local to one another, or may be located physically distant to one another (in different offices, buildings, cities, countries, 20 etc.).

The database system 514 includes a sequential log 518 and a controller 516 which controls access to the sequential log 518. The sequential log 518 includes a plurality of physical memory blocks (for simplicity, only six blocks 520A-520F are shown in FIG. 5). Preferably, the size of each block 520 is the same as the size of the internal buffers 504, 508, 512. In the example of FIG. 5, the blocks 520 and the internal buffers 504, 508, 512 can each store three data

A 30

records.

Each of the computer systems 502, 506, 510 is preferably implemented using a computer 602 having a block diagram as shown in FIG. 6. The computer 602 includes one or more processors 604 connected to a bus 606. The computer 602 3J also includes a main memory 608, preferably random access memory (RAM), and a secondary memory 610 that includes, for example, a disk drive 612.

The computer 602 further includes a network interface 618 that couples the computer 602 to a communication 40 network 620. Preferably, the database system 514 is connected to this communication network 620, such that the first, second, and Nth computer systems 502, 506, 510 communicate with the database system 514 via the communication network 620 (other well known means for connect- 45 ing the computer systems 502, 506, 510 to the database system 514 can alternatively be used).

A suitable form of the computer 602 is the well known RISC System/6000 family of computers produced by IBM. Alternatively, the computer 602 is the well known System/ 50 390 family of computers produced by IBM. It should be understood, however, that other computers could alternatively be used without departing from the scope and spirit of the present invention.

The database system 514 may also be implemented using 55 a computer 602 as shown in FIG. 6. In this case, the sequential log 518 is stored in the main memory 608 and/or the secondary memory 610. The controller 516 preferably represents software which (when executing) resides in main memory 608 (alternatively, the controller 516 is imple- 60 mented primarily using hardware, as a hardware state machine, for example). (The computer program representing the controller 516 may be stored in a floppy disk or some other removable storage unit 616, which is read by the disk drive or some other storage device 612 and then transferred 65 to the main memory 608.) The controller 516, when executed, enables the computer 602 to perform the features

6

of the present invention as discussed herein. Thus, the controller 516 represents a controller of the computer 602.

The operation of the present invention shall now be described.

FIG. 7 is a flowchart 702 illustrating the operation of each of the computer systems 502, 506, 510 when storing a new data record in the associated internal buffer 504, 508, 512, respectively. For illustrative purposes, flowchart 702 shall be described with reference to the first computer system 502 (although the other computer systems 506, 510 operate in the same way). Flowchart 702 begins with step 704, where control immediately passes to step 706.

In step 706, the first computer system 502 creates a new data record. The new data record may be created in response to a command from a user. For example, in a banking situation, a new data record is created each time a customer performs a transaction (such as depositing, transferring, and withdrawing funds). The structure of the data record and the information stored in the data record are implementation dependent, and will be apparent to persons skilled in the relevant art.

In step 708, the first computer system 502 obtains the current date/time stamp (i.e., the current date and time). Preferably, the first computer system 502 performs step 708 by accessing and obtaining the current date/time stamp from a clock device, such as coordinate clock 522. The coordinate clock 522 keeps track of the current date and time, and makes this information available to the computer systems 502, 506, 510 upon request (the coordinate clock 522 may be connected to the computer systems 502, 506, 510 via a network, for example). Coordinate clocks are well known devices and, thus, the structure and operation of the coordinate clock 522 will be apparent to persons skilled in the art.

Alternatively, the first computer system 502 obtains the current date/time stamp in step 708 by acquiring the current date and time from its internal clock (not shown), which is synchronized with the internal clocks (not shown) in the other computer systems 506,510. According to this alternate embodiment, the internal clocks in the computer systems 502, 506, 510 are periodically synchronized to one another using well known procedures (for example, by having the computer systems 502, 506, 510 update their internal clocks using the current date and time provided by the coordinate clock 522).

In step 710, the first computer system 502 stores the data record and the date/time stamp in the next available (i.e., free) memory location of its internal buffer 504 (the data record could be considered to include both the data and the date/time stamp). Thus, according to the present invention, a date/time stamp is associated with each data record. When its internal buffer 504 becomes full (in the example of FIG. 5, the buffer 504 becomes full after three data records have been stored in it), the first computer system 502 transfers the content of the internal buffer 504 to the database system 514. The operation of flowchart 702 is complete after step 710 is performed, as indicated by step 712.

FIG. 8 is a flowchart 802 representing the operation of the database system 514 when it receives a data block from one of the computer systems 502, 506, 510. A data block preferably represents the contents of an internal buffer 504, 508, or 512 (i.e., three data records in the example of FIG. 5). Flowchart 802 begins with step 804, where control immediately passes to step 806.

In step 806, the database system 514 receives a data block from one of the computer systems 502, 506, 510.