Cluster computing using special purpose microprocessors

• US 8,676,877 B2
• Filed: 03/16/2012
• Issued: 03/18/2014
• Est. Priority Date: 06/13/2006
• Status: Active Grant

• Current Assignee: Advanced Cluster Systems, Inc.
• Original Assignee: Advanced Cluster Systems, Inc.
• Inventors: Tannenbaum, Zvi, Dauger, Dean E.
• Primary Examiner(s): DONAGHUE, LARRY D
• Application Number:
  US13/423,063
• Publication Number:
  US 20130097406A1
• Time in Patent Office: 732 Days
• Field of Search: 709/201, 719/320, 717/177
• US Class Current: 709/201
• CPC Class Codes: G06F 15/76 Architectures of general purpose stored program computers with program plugboard G06F15/08; multicomputers G06F15/16
• CPC Class Codes: G06F 9/5072 Grid computing
• CPC Class Codes: G06F 9/54 Interprogram communication
• CPC Class Codes: H04L 41/04 Network management architectures or arrangements

Abstract

In some embodiments, a computer cluster system comprises a plurality of nodes and a software package comprising a user interface and a kernel for interpreting program code instructions. In certain embodiments, a cluster node module is configured to communicate with the kernel and other cluster node modules. The cluster node module can accept instructions from the user interface and can interpret at least some of the instructions such that several cluster node modules in communication with one another and with a kernel can act as a computer cluster.

Claims

1. A system for performing an instruction received from a front end by executing commands on one or more special purpose microprocessors, the system comprising:

a plurality of nodes, wherein each node is configured to access a computer-readable memory system comprising program code for a single-node kernel module, and wherein each single-node kernel module is configured to interpret instructions received by the single-node kernel module into commands that are executable by a special purpose microprocessor;

a plurality of cluster node modules, wherein each cluster node module is stored in a computer-readable memory system and configured to communicate with a single-node kernel and with one or more other cluster node modules, to accept instructions, and to interpret at least some of the instructions such that the plurality of cluster node modules communicate with one another in order to act as a cluster in executing commands using one or more hardware processors; and

a communications system configured to connect the plurality of nodes;

wherein the plurality of cluster node modules cooperate to interpret and translate, as needed, the instruction for execution by a plurality of single-node kernel modules, and wherein at least one of the plurality of cluster node modules returns a result to the front end.

2. The system of claim 1, wherein the special purpose microprocessor comprises a digital signal processor.

3. The system of claim 1, wherein the plurality of nodes are organized into two or more groups of node subsets.

4. The system of claim 3, wherein at least one of the two or more groups of node subsets exchange data with the special purpose microprocessor.

5. The system of claim 1, wherein the special purpose microprocessor comprises multiple processor cores.

6. The system of claim 1, wherein at least one of the plurality of cluster node modules resides in processor cache memory.

7. The system of claim 1, wherein each single-node kernel modules resides in processor cache memory.

8. A system for performing an instruction received from a front end by executing commands on one or more hardware processors, the system comprising:

a plurality of nodes, wherein each node is configured to access a computer-readable memory system comprising program code for a single-node kernel module, and wherein each single-node kernel module is configured to interpret instructions received by the single-node kernel module into commands that are executable by a hardware processor, wherein the commands are configured to perform computations one or more elements from a list of elements;

a plurality of cluster node modules, wherein each cluster node module is stored in a computer-readable memory system and configured to communicate with a single-node kernel and with one or more other cluster node modules, to accept instructions, and to interpret at least some of the instructions such that the plurality of cluster node modules communicate with one another in order to act as a cluster in performing computations on the list of elements; and

a communications system configured to connect the plurality of nodes;

wherein the plurality of cluster node modules cooperate to interpret and translate, as needed, the instruction for execution by a plurality of single-node kernel modules, wherein the list of elements is partitioned for execution on the plurality of nodes, and wherein at least one of the plurality of cluster node modules returns a result to the front end.

9. The system of claim 8, wherein one or more elements of the list of elements migrates to a different node.

10. A method of performing an instruction received from a front end by executing commands on one or more special purpose microprocessors, the method comprising:

communicating an instruction from a front end to one or more cluster node modules connected to one another by a communications system;

for each of the one or more cluster node modules, communicating a message based on the instruction to a respective kernel module associated with the cluster node module, wherein the respective kernel module is configured to interpret the message into commands that are executable by a special purpose microprocessor;

for each of the one or more cluster node modules, receiving a result from the respective kernel module associated with the cluster node module; and

for at least one of the one or more cluster node modules, responding to messages from other cluster node modules.

11. The method of claim 10, wherein communicating a message based on the command to a respective kernel module associated with the cluster node module comprises communicating a specially identified message to the respective kernel module.

12. The method of claim 10, wherein responding to messages from other cluster node modules comprises:

forwarding messages from the other cluster node modules to a received message queue;

matching data from each entry in a message receiving queue with entries in the received message queue;

combining data from the message receiving queue with matching data in the received message queue; and

marking the matching data as completed.

13. The method of claim 10, wherein communicating a command from at least one of a user interface or a script to one or more cluster node modules within the computer cluster comprises communicating an instruction from a Mathematica Front End to a first cluster node module, wherein the first cluster node module forwards the instruction to other cluster node modules running on the computer cluster.

14. The method of claim 10, wherein each of the one or more cluster node modules communicates with a respective kernel module using MathLink.