Accessing remote data objects in a distributed memory environment using parallel address locations at each local memory to reference a same data object

• US 5,655,101 A
• Filed: 05/27/1994
• Issued: 08/05/1997
• Est. Priority Date: 06/01/1993
• Status: Expired due to Term

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: O'Farrell, William G., Arjomandi, Eshrat
• Primary Examiner(s): Chan, Eddie P.
• Assistant Examiner(s): Ellis, Kevin L.
• Application Number:
  US08/250,202
• Time in Patent Office: 1,166 Days
• Field of Search: 395/474, 395/475, 395/497.01, 395/412, 395/200.01, 395/200.03, 395/200.09, 395/200.16
• US Class Current: 711/148
• CPC Class Codes: G06F 12/0284 Multiple user address space allocation, e.g. using different base addresses interprocessor communication G06F15/163
• CPC Class Codes: G06F 9/465 Distributed object oriented systems remote method invocation [RMI] G06F9/548

Abstract

A mechanism and method for accessing remote data objects in a distributed memory environment is disclosed. In the distributed memory environment, a number of parallel processors are remote from each other and each has memory storage capacity with parallel address locations. For each data object stored in the local memory of a processor, that processor stores a variable at a specific location pointing to the address of the data object in its local memory. At the same (parallel) location in the local memories of all remote parallel processors, a variable pointing to the home processor for the data object is stored. In order to access the data object, it is only necessary to locate the identifier variable in the processor processing a program pointing to the processor having the data object in storage. Because of the parallel address space, the location will automatically point directly to the same address location in the home storage processor, and this address location contains the address pointing to the actual location of the data object in the local storage of that processor. Alternate means for updating the variable information on data object storage is also provided. In one system, changes are broadcast throughout the parallel processor environment while in the other system, changes are merely retained on the home processor or the home processor and one remote parallel processor responsible for recording changes in that region of the distributed memory environment. When accessed by another remote parallel processor, the change information will be transmitted.

Claims

1. A composite mechanism for accessing remote data objects in a distributed memory environment having a plurality of parallel processors remote from each other, each processor having a local memory with parallel address locations, wherein for each data object having a storage location in the local memory of a first processor, the mechanism comprises:

a first variable storage repository at a first address location in the local memory of the first processor pointing to said object'"'"'s storage location in the memory of the first processor; and

a second variable storage repository at the first address location in the local memory of at least one remote parallel processor pointing to the first processor, whereby the first address location in the local memory of the first processor and the first address location in the local memory of the at least one remote parallel processor have a same corresponding address location.

2. A composite mechanism, according to claim 1, wherein the second variable storage repository is located in the local memory of every remote parallel processor in said environment.

3. A composite mechanism, according to claim 1, wherein the second variable storage repository is located in one remote parallel processor, and further comprising a third variable storage repository at the first address location in the local memory of remaining remote parallel processors pointing to said one remote parallel processor.

4. A composite mechanism, according to claim 1, further comprising relay means for changing variable storage repositories in response to data object migration, said relay means comprising:

means for creating and destroying first and second variable repositories; and

means for replacing first and second variable repositories.

5. A composite mechanism, according to claim 4, wherein the relay means further comprise means for broadcasting changes in variable storage repositories to all remote parallel processors in the environment.

6. A composite mechanism, according to claim 4, further comprising means for transmitting changes in variable storage repositories to at least one remote parallel processor in the environment.

7. A composite mechanism, according to claim 1, further comprising an interface mechanism for redirecting local calls from a calling data object in the local memory of the first processor to a remote processor.

8. A composite mechanism, according to claim 7, wherein the interface comprises a protocol function table indexed to execution data for redirecting local calls to remote processors, and further comprising:

a virtual function table indexed to execution data for virtual functions;

means for constructing the calling data object as an object with virtual functions;

means for indexing the calling data object to the protocol function table; and

means for indexing the protocol function table to the virtual function table.

9. A mechanism for accessing remote data objects in a distributed memory environment having a plurality of parallel processors remote from each other, each processor having a local memory with parallel address locations and each processor adapted to be a home processor storing data objects in said local memory, comprising:

for each home processor;

a first repository storage segment in the local memory of the home processor at an address location, said first repository storage segment being adapted to contain variables pointing to the data objects concurrently stored in the local memory of said home processor;

a second repository storage segment in the local memory of all remote parallel processors at the address location, said second repository storage segment being adapted to contain identifiers for the data objects concurrently stored in the local memory of said home processor, whereby the address location in the local memory of the home processor and the address location in the local memory of the remote parallel processors have a same corresponding location; and

relay means for changing said first and second repository segments in response to a change in data object storage in said local memory of the home processor.

10. A mechanism, according to claim 9, wherein said second repository storage segments are adapted to contain variables pointing to the home processor.

11. A mechanism, according to claim 9, wherein said second repository storage segments are adapted to contain a value indicating the home processor.

12. The mechanism of claim 9 wherein the address location of the second repository storage segment indicates the home processor.

13. A method of accessing remote data objects in a distributed memory environment between remote parallel processors, each parallel processor having a local memory and being adapted to be a home processor storing at least one data object in said local memory, comprising the computer implemented steps of:

storing an address variable for a data object at a location in the local memory of the home processor;

storing an identifier variable for the data object at the location in the local memory of at least one remote parallel processor, whereby the location in the local memory of the home processor and the location in the local memory of the at least one remote parallel processor have a same corresponding location;

locating said identifier variable for accessing said address variable; and

locating said address variable for accessing said data object in the local memory of the home processor.

14. A method, according to claim 13, further comprising the step of:

changing said address variable in response to migration of the data object in the local memory of the home processor.

15. A method, according to claim 13, further comprising the step of:

converting said address variable to an identifier variable at a location in local memory pointing to a remote parallel processor; and

creating an address variable at the same location in the local memory of the remote parallel processor for an address in the local memory of said remote parallel processor, in response to migration of the data object outside the memory of the home processor to the address in the local memory of said remote processor.

16. A method, according to claim 15, further comprising the step of:

transmitting said identifier variable pointing to a remote parallel processor from the home processor to all remote parallel processors in the environment.

17. A method, according to claim 15, wherein the step of locating said identifier variable further comprises the step of transmitting said identifier variable to a remote parallel processor from the home processor.

18. A method, according to claim 13, further comprising the steps of:

constructing a calling data object in dynamic memory storage in the local memory of a remote processor;

creating a generic operation for redirecting a call from a data object to a remote processor; and

linking the calling data object with the generic operation.

19. A method, according to claim 18, wherein the step of constructing the calling data object comprises the steps of:

declaring the calling data object as having virtual functions, and

creating a first pointer in the calling data object indexing virtual functions in the calling data object to a virtual function table; and

wherein the step of linking the calling data object with the generic operation comprises the steps of;

storing the first pointer in the calling data object as a variable, and

replacing the first pointer with a second pointer indexing the virtual functions in the calling data object to a protocol function table indexed to the generic operation.