Technology for integrated computaion and communication; TICC

• US 20040044794A1
• Filed: 10/07/2002
• Published: 03/04/2004
• Est. Priority Date: 10/15/2001
• Status: Active Grant

• Current Assignee: EDSS Incorporated
• Original Assignee: EDSS Incorporated
• Inventors: Srinivasan, Chitoor V.
• Granted Patent: US 7,210,145 B2
• Time in Patent Office: Days
• US Class Current: 709/248
• CPC Class Codes: G06F 9/544 Buffers; Shared memory; Pipes

Abstract

TICC manages asynchronous communications among groups of concurrent (parallel) processes (Objects) in multiprocessors. It dedicates one processor to function as Communications Processor, C. TICC defines a Causal Communication programming Language, called CCL, whose statements may appear intermixed with statements of any conventional programming language. TICC defines methods for compiling CCL statements in these mixed language programs into sequences of protocols which are executed by C, in parallel with on going computations guaranteeing (1) group-to-group loss less, buffer less, self-synchronizing asynchronous data transfers; (2) more than a hundred fold reduction in communication latencies; (3) Dynamic Flexibility to monitor, repair, reconfigure and update software objects without service interruption; and (4) protection and security of all data, distributed to or collected from communicating objects, based on an Agreement Protocol.

Claims

1. A method for the operation of a computer-based multiprocessing, with two or more processors, system composed of software objects, the objects being programs with TICC-ports, or canonical representations of such programs with TICC-ports, objects being capable of operating concurrently or in parallel, and communicating with each other asynchronously through their TICC-ports, via specially established TICC-pathways containing TICC-agents, at least one of the processors being dedicated as the communications processor, the method comprising the steps of

Dynamically establishing TICC-pathways between TICC-ports belonging to distinct groups of objects and modifying existing TICC-pathways, without service interruption and without loss of data flowing through TICC-pathways, while the multiprocessing system is operating;

Specifying group-to-group communications in the program of each object, as necessary, in a Causal Communication Language (CCL), which may be freely intermixed with any conventional programming language;

Compiling CCL statements that appear in programs into sequences of communication protocols, which are efficiently executed in parallel with the programs by a dedicated communications processor, in order to deliver data asynchronously from one group of objects to another group, via protected and shared memories in already established TICC-pathways, in a self-synchronizing, loss less, buffer free manner, with extremely low latencies;

2. The method as recited in claim 1 further including the steps for

Enforcing application system security, while TICC-pathways are being established between TICC-ports, and reporting on attempts made by objects, if any, to establish TICC-pathway connections that violate application system security;

Tuning a TICC-agent to a group of TICC-ports, each belonging to a distinct object, so that agent may distribute persistent and protected data to each port it is tuned to in each cycle of computation, and ports may deliver the data they receive from their agents to their respective parent objects;

Interlocking tuned TICC-agent and TICC-ports, in each cycle of computation to ensure that all computations performed by the respective parent objects of the TICC-ports in each cycle, on data received from their ports, are completed, and the agent advances to the next cycle from the present cycle only after the agent has received agreement to do so from all the objects, and no object ever receives new input data while it is still busy performing computations on its current inputs.

3. The method as recited in claim 2 further including the steps for

Dynamically enforcing application system security on each message sent/received by each TICC-port, and dynamically monitoring security status so that no TICC-agent would distribute (or receive) data to (from) any of the TICC-ports it is tuned to, in violation of application system security:

Installing protocols for dynamically monitoring application system performance, at every TICC-port, reporting on time-outs, poor performance, detecting system failures and taking corrective measures.

6. Methods as recited in claims 1 through 5 above, implemented in hardware or software, including steps for

Using TICC agents, ports, clock-rings and watch-rings, or their equivalents, compiling causal TICC communications statements, or their equivalents, to sequences of protocols and presenting them to a dedicated Communications Processor, executing the protocols in parallel with computations to deliver data to their intended recipients and enforce application system security, and using a dedicated processor to dynamically self-monitor, diagnose and repair application systems, in parallel with computations being performed by the application.

4. A method for the operation of a computer-based multiprocessing system composed of a plurality of objects operating concurrently or in parallel, the objects being programs with TICC-ponts, or canonical representations of such programs with TICC-ports, the method comprising the steps of

Dynamically attaching monitor agents and monitor networks to existing TICC-pathways to observe and change data flowing through TICC-pathways, without service interruption and without loss of any of the on going data transfers, attaching such monitor ports to any point on existing TICC-pathways except inside objects, agents and ports.

5. A method as recited in claim 4 further including the steps for

Dynamically installing TICC-pathways with agents and objects that are needed to temporarily suspend computations on any selected set of application objects;

Dynamically installing objects consisting of programs needed to reconfigure an application system according to specified reconfiguration criteria, without service interruption and without loss of any of the benefits already obtained from computations done up to the time of reconfiguration, if such reconfiguration was consistent with correct operation of an application system;

Dynamically installing TICC-pathways and networks of objects needed to test new versions of application objects (or application subsystems) in exactly the same context in which their older versions function, in parallel with the older versions, and replace the older versions with the new versions, after testing has been successfully completed, all without service interruption, and thus making possible dynamic evolution of application systems