System partitioning for massively parallel processors
First Claim
1. In a distributed parallel processing environment having a plurality of nodes connected to one another via a network in such way as to form a single system unit, a method of logically partitioning said processing environment into at least two isolated sub-environments, said method comprising the steps of:
- controlling all communications through a central control element connected to said nodes via said network;
storing all system data in a system data repository located in said central control element;
identifying each node as either belonging to a first sub-environment or a second sub-environment; and
organizing data deposited in said system data repository into system object classes providing data globally accessible to all nodes, and partitioned object classes providing data that is exclusive to each sub-environment.
1 Assignment
0 Petitions
Accused Products
Abstract
In accordance with a preferred embodiment of the present invention, a method and apparatus is disclosed for partitioning a massively distributed parallel processor into at least two isolated sub-environments. The parallel processing environment is comprised of a plurality of nodes connected to one another via a network in such a way as to form a single system unit initially. A central control element is provided for controlling and maintaining all communications and status information between nodes. The nodes and the central control element are in processing communication with one another via the network. All system data is deposited in a system data repository located in the central control element. Once the environment is partitioned, the system data stored in this repository is organized into system or partitioned classes, with the partitioned classes only being accessible selectively to nodes residing in one environment or another.
-
Citations
33 Claims
-
1. In a distributed parallel processing environment having a plurality of nodes connected to one another via a network in such way as to form a single system unit, a method of logically partitioning said processing environment into at least two isolated sub-environments, said method comprising the steps of:
-
controlling all communications through a central control element connected to said nodes via said network; storing all system data in a system data repository located in said central control element; identifying each node as either belonging to a first sub-environment or a second sub-environment; and organizing data deposited in said system data repository into system object classes providing data globally accessible to all nodes, and partitioned object classes providing data that is exclusive to each sub-environment. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24)
-
-
25. A massively parallel processing environment partitioned into sub-environments and having a plurality of nodes connected via a network in such a way as to initially create a single system unit, said processing environment comprising:
-
a central control element for controlling all communications between said nodes and for maintaining configuration and status information, said central control element being in processing communications with said nodes via said network; at least two partitioned sub-environments each having exclusive ownership of a subset of nodes in a manner that no two nodes in said system unit environment belong to said same sub-environment; a system data repository located in said central control element for storing system data, said data in said repository organized into a plurality of object classes; and said object classes being divided into system and partitioned classes, said system classes having data targeted to be globally available to all nodes on both sub-environments and said data in said partitioned classes being exclusively accessible only to nodes of one sub-environment or another. - View Dependent Claims (26, 27, 28, 29, 30, 31, 32, 33)
-
Specification