Method and system for improving performance and scalability of applications that utilize a flow-based-programming methodology

• US 20050289505A1
• Filed: 06/25/2004
• Published: 12/29/2005
• Est. Priority Date: 06/25/2004
• Status: Abandoned Application

• Current Assignee: Stanley Norman Williams
• Original Assignee: Stanley Norman Williams
• Inventors: Williams, Stanley Norman
• Application Number:
  US10/876,321
• Publication Number:
  US 20050289505A1
• Time in Patent Office: Days
• US Class Current: 717/103
• CPC Class Codes: G06F 9/485 Task life-cycle, e.g. stopping, restarting, resuming execution G06F9/4881 takes precedence
• CPC Class Codes: G06F 9/4881 Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
• CPC Class Codes: G06F 9/544 Buffers; Shared memory; Pipes

Abstract

A method, system, apparatus, and computer program product is presented for improving the execution performance of flow-based-program (FBP) programs and improving the execution performance further on systems with additional processing resources (scalability). A FBP supervisor is inserted as the initial executable program, which program will interrogate the features of the operating system upon which it is executing including but not limited to number of processors, memory capacity, auxiliary memory capacity (paging dataset size), and networking capabilities. The supervisor will create an optimum number of processing environments (e.g. threads in a Windows environment) to service the user FBP application. The supervisor will further expose other services to the FBP application which improve the concurrent execution of the work granules (processes) within that FBP application. The supervisor further improves the generation and logging of messages through structured message libraries which are extended to the application programmer. The overall supervisor design maximizes concurrency, eliminates unnecessary work, and offers services so a process should suspend rather than block.

Claims

1. A method, data processing system, and computer program product for improving the services available to and execution performance of a FBP application in a data processing system, wherein the method offers services to the application that permit more work to be performed with less overhead, eliminate circumstances where the computer waits for services while other work may be performed, and permit dynamic adjustment in execution priorities for time-dependant processes.

2. The method, data processing system, and computer program product of claim 1, wherein with no changes in the user application, it will dynamically recognize the number of processors available in the executing system establishing a dispatching environment for each of those available processors.

3. The method, data processing system, and computer program product of claim 1, wherein an application execution may be distributed across multiple execution environments. This is achieved through establishing connections with different systems and exchanging services including but not limited to application networks, application processes and execution libraries, message libraries, and data packets.

4. The method, data processing system, and computer program product of claim 1, wherein data Input/Output (I/O) services are offered to the application.

5. The method, data processing system, and computer program product of claim 1, wherein asynchronous I/O operations are simulated on data processing systems that do not support asynchronous I/O operations.

6. The method, data processing system, and computer program product of claim 1, wherein pure asynchronous I/O operations are performed on data processing systems that support asynchronous I/O operations

7. The method, data processing system, and computer program product of claim 1, wherein ‘

virtual buffers’

are utilized, where appropriate, to reduce system overhead.

8. The method, data processing system, and computer program product of claim 1, wherein the optimal I/O methods for the executing data processing system are selected without change to the application requesting that I/O operation.

9. The method, data processing system, and computer program product of claim 1, wherein the application may request and receive concurrent I/O operations without intervening blocking.

10. The method, data processing system, and computer program product of claim 1, wherein application I/O requests may specify a control field (event) which will be changed when the I/O operation completes (posted), and may be waited upon until it is posted.

11. The method, data processing system, and computer program product of claim 1, wherein I/O operations may be requested on temporary files which are memory backed rather than requiring physical I/O operations which require more clock time and data processing system overhead.

12. The method, data processing system, and computer program product of claim 1, wherein the application process may prepare for (open) and terminate those preparations (close) I/O operations to a data processing system file, overriding default settings.

13. The method, data processing system, and computer program product of claim 1, wherein the application process may request I/O services including, but not limited to requesting data from a file (READ), Writing data to a file (WRITE), checking the status of an operations (CHECK), and repositioning for the next operation (SEEK).

14. The method, data processing system, and computer program product of claim 1, wherein one or more routines may be specified (NoWorkList) to receive control when all processes are waiting for something else to happen (deadlock). This routine may wait for some external action, such as a message down a pipe, or may inspect the operational network for full port conditions and insert corrective processes, or may execute a background function to utilize the idle processor.

15. The method, data processing system, and computer program product of claim 1, wherein a process may request immediate entry to the NoWorkList to provide application wide synchronization.

16. The method, data processing system, and computer program product of claim 1, wherein a special dynamic buffering process may be inserted into a network preventing deadlock conditions caused by excessive information packets (IPs). The dynamic buffering process will, progressively, save excessive IPs in memory until the downstream process starts accepting IPs, compress in-memory IPs into fixed sized memory chunks, write compressed chunks to a temporary file or the paging dataset as defaulted, reverse the previous functions supplying IPs in the original order.

17. The method, data processing system, and computer program product of claim 1, wherein a real-time, time dependant or other process, may have its priority boosted or dropped to meet service level agreements. A process may request its own priority, or the priority of another process, to be set higher, set lower, or set to increase steadily as it takes longer to complete.

18. The method, data processing system, and computer program product of claim 1, wherein a network which is named but not defined in an application may be loaded dynamically at execution time (dynamic network). The dynamic network may carry its own services such as message library, functions, link libraries, and lower level networks.

19. The method, data processing system, and computer program product of claim 1, wherein free IPs are maintained in pools of preset sizes for fast allocation and return. Each pooled IP carries a data overrun flag, which when found altered, triggers an error condition.

20. The method, data processing system, and computer program product of claim 1, wherein a message service permits messages from many sources to be combined into a single log.

21. The method, data processing system, and computer program product of claim 1, wherein messages support substitution. The calling process, including called services, may request a message by message-number, and supply values for substitution into the base message text.

22. The method, data processing system, and computer program product of claim 1, wherein each message may specify the order of substitutions and may specify the same value to appear multiple times in the message text.

23. The method, data processing system, and computer program product of claim 1, wherein multiple like named message libraries may be specified and selected by a language code. A message in one language will likely specify different text and order of substitution.

24. The method, data processing system, and computer program product of claim 1, wherein a message library may be associated with a unique process, network, or distribution library (DLL). Multiple message libraries with the same name are supported and are searched in order of closeness to the running process, that is process first, then contained network, then contained DLL, then the default system library. A message not found in one library may be located in a lower level library.

25. The method, data processing system, and computer program product of claim 1, wherein a process may monitor a directory or file changes through non-blocking operations. Changes are reported in a stream of IPs to an open port on the calling process. The process specifies the level of monitoring desired.

26. The method, data processing system, and computer program product of claim 1, wherein the method offers services to the application that permit data to be pushed from one process to another with reduced overhead, referred in this invention as a turbo port and a turbo exit.

27. The method, data processing system, and computer program product of claim 1, wherein a process may send data to an output port in a standard manner without knowing the processing mode of the receiving input port. I.e. turbo or normal.

28. The method, data processing system, and computer program product of claim 1, wherein a process may “

Freeze”

the input port thereby preventing any upstream process from pushing additional data into the turbo exit. The data is queued pending further action.

29. The method, data processing system, and computer program product of claim 1, wherein a process may “

Thaw”

a previously frozen port. Any data pending due to a freeze will be pushed to the turbo exit followed by, within the same call or subsequent calls, any new data arriving at the input port.

30. The method, data processing system, and computer program product of claim 1, wherein a process may “

Hold”

any data already pushed into a turbo port. This data is returned to the input port for future processing.

31. The method, data processing system, and computer program product of claim 1, wherein a process may query an output port to determine the “

Limit”

or maximum number of Information Packets (IPs) that will be accepted before the port suspends or sets an event.