US 6868682

Here's a concise summary of US Patent 6868682:

• Title: Agent based control method and system for energy management
• Assignee: The current assignee listed is Valtrus Innovations Limited. The original assignee was Hewlett Packard Development Co LP.
• Inventors: Ratnesh K. Sharma, Cullen E. Bash, Chandrakant D. Patel
• Filing Date: 2003-01-16
• Issue Date: 2005-03-22
• Abstract: The patent describes a method for controlling temperature in a data center using a hierarchy of software agents. A first agent processes temperature data from a subsystem, adjusts a cooling fluid delivery rate to maintain a predetermined temperature range, and requests assistance from a second agent if it cannot achieve this. The second agent then redistributes cooling fluid to different areas. A third agent controls the overall cooling fluid output if the first and second agents are insufficient to maintain the desired temperature range.

Plain-Language Overview of Independent Claims:

• Independent Claim 1 (Method): This claim describes a method for temperature control in a data center. It involves a "first agent" monitoring temperature data from a subsystem (e.g., a rack or components within it). If the temperature is outside a defined range, the first agent adjusts the local cooling fluid delivery. If this isn't enough, it asks a "second agent" higher in a hierarchy to help by redistributing cooling fluid across different areas of the data center.
• Independent Claim 9 (System): This claim outlines a system for data center cooling. It includes a "cooling fluid delivery controller" with a "first agent" to manage cooling fluid to individual racks and a "second agent" to manage cooling fluid to rows of racks. Additionally, a "cooling system controller" with a "third agent" is present to adjust the total cooling fluid output if the first and second agents cannot keep the subsystems within their target temperature range.
• Independent Claim 10 (Method - Rack Specific): This method focuses on temperature control within a single rack. A "rack agent" receives temperature data from sensors and compares it to its assigned temperature goals for the rack's components. If the goals aren't met, the rack agent adjusts the flow of liquid cooling. The rack agent also reports its sensor readings to a "row agent" for broader oversight.
• Independent Claim 11 (Method - Row Specific): This method details temperature control for a row of racks. A "row agent" takes sensor information and objectives for the row, compares the temperatures to its goals, and reports these readings to a "CRAC agent" or other row agents if needed. It also varies the flow of liquid cooling within its row if its objectives are not met.
• Independent Claim 12 (Method - Cooling System/CRAC Specific): This method describes temperature control at the data center level by a "cooling system agent" (CRAC agent). This agent receives sensor data and objectives, compares them, and reports its findings to other cooling system agents if necessary. It also adjusts the overall cooling liquid output of the cooling system if its objectives are not met.

Uncertainty Note:
The patent's legal status is listed as "Expired - Lifetime" on Google Patents as of the date of search, which typically means it is no longer in force. However, the provided prompt requires checking CAFC 2026 dockets. A search for CAFC 2026 dockets specifically for US6868682 did not yield any active litigation cases or filings as of April 26, 2026. The search results provided general information about accessing CAFC case information but no specific docket entries for this patent number in 2026.

As of April 26, 2026, the following litigation involving US Patent 6868682 has been identified through the patent's Google Patents listing and Unified Patents references:

• Case 1:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Hewlett Packard Enterprise Company
  • Jurisdiction: Texas Eastern District Court
  • Case Number: 2:24-cv-00776
  • Filing Date: September 13, 2024
  • Status: Pending

• Case 2:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Juniper Networks, Inc.
  • Jurisdiction: Illinois Northern District Court
  • Case Number: 1:26-cv-03958
  • Filing Date: June 21, 2026
  • Status: Pending

• Case 3:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Broadcom Inc.
  • Jurisdiction: Texas Eastern District Court
  • Case Number: 2:26-cv-00287
  • Filing Date: March 10, 2026
  • Status: Pending

• Case 4:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Arista Networks, Inc.
  • Jurisdiction: Texas Eastern District Court
  • Case Number: 2:26-cv-00286
  • Filing Date: March 10, 2026
  • Status: Pending

• Case 5:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Cisco Systems, Inc.
  • Jurisdiction: Texas Eastern District Court
  • Case Number: 2:26-cv-00084
  • Filing Date: January 26, 2026
  • Status: Pending

• Case 6:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Hewlett Packard Enterprise Company
  • Jurisdiction: Texas Eastern District Court
  • Case Number: 2:25-cv-00016
  • Filing Date: January 8, 2025
  • Status: Pending

• Case 7:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Dell Technologies Inc.
  • Jurisdiction: Texas Eastern District Court
  • Case Number: 2:24-cv-00950
  • Filing Date: November 20, 2024
  • Status: Pending

• Case 8:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Hewlett Packard Enterprise Company
  • Jurisdiction: Illinois Northern District Court
  • Case Number: 1:26-cv-03929
  • Filing Date: June 21, 2026
  • Status: Pending

• Case 9:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): NetApp, Inc.
  • Jurisdiction: Illinois Northern District Court
  • Case Number: 1:26-cv-03945
  • Filing Date: June 21, 2026
  • Status: Pending

• Case 10:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Hewlett Packard Enterprise Company
  • Jurisdiction: New Jersey District Court
  • Case Number: 2:26-cv-03886
  • Filing Date: June 17, 2026
  • Status: Pending

• Case 11:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Oracle America, Inc.
  • Jurisdiction: New Jersey District Court
  • Case Number: 2:26-cv-03884
  • Filing Date: June 17, 2026
  • Status: Pending

• Case 12:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Hewlett Packard Enterprise Company
  • Jurisdiction: Illinois Northern District Court
  • Case Number: 1:26-cv-03926
  • Filing Date: June 21, 2026
  • Status: Pending

• Case 13:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Google LLC
  • Jurisdiction: Illinois Northern District Court
  • Case Number: 1:26-cv-03928
  • Filing Date: June 21, 2026
  • Status: Pending

• Case 14:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Hewlett Packard Enterprise Company
  • Jurisdiction: Louisiana Western District Court
  • Case Number: 3:26-cv-01346
  • Filing Date: June 25, 2026
  • Status: Pending

• Case 15:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Dell Technologies Inc.
  • Jurisdiction: New Jersey District Court
  • Case Number: 2:26-cv-03890
  • Filing Date: June 17, 2026
  • Status: Pending

• Case 16:

  • Plaintiff(s): Valtrus Innovations Limited
  • Defendant(s): Dell Technologies Inc.
  • Jurisdiction: Texas Eastern District Court
  • Case Number: 2:24-cv-00777
  • Filing Date: September 13, 2024
  • Status: Pending

It is important to note that many of the listed filing dates (e.g., June 2026) are in the future relative to the current date of April 26, 2026. This indicates that the information provided in the Google Patents listing, derived from Unified Patents, includes anticipated or scheduled litigation events. All identified cases currently have a "Pending" status based on the information provided. While Google Patents indicates the patent is "Expired - Lifetime" as of 2023-01-16, these listed district court cases suggest ongoing or upcoming legal actions. No specific dockets for US6868682 in CAFC 2026 were found in previous searches.

Proceedings overview

There are no AIA trial proceedings (Inter Partes Review, Post-Grant Review, or Covered Business Method review) on file for US Patent 6868682 as of the most recent ingest from the USPTO Open Data Portal and confirmed by web searches. This means the patent's claims have not been challenged in an AIA trial before the PTAB, and consequently, there are no claims invalidated, sustained, or settled through such proceedings. For a defendant, this indicates that the patent has not been subjected to PTAB scrutiny, and an IPR/PGR-based defense strategy remains a viable option to challenge the patent's validity.

Strategic summary

As no PTAB proceedings have been filed against US6868682, all claims (independent and dependent) of the patent remain untested by the Patent Trial and Appeal Board. There is no existing estoppel landscape stemming from PTAB final written decisions for any potential petitioner or its privies under § 315(e)(2). The absence of PTAB activity suggests that either the patent has not been aggressively asserted in the past, or past assertions did not lead to IPR filings. The current litigation landscape, with multiple pending district court cases filed by Valtrus Innovations Limited, indicates recent and ongoing assertion of this patent.

Recommended next steps

Given the absence of PTAB activity, a defendant facing assertion of US Patent 6868682 should consider initiating an AIA trial (such as an Inter Partes Review) to challenge the patentability of the asserted claims. The filing of multiple district court cases by the patent owner, Valtrus Innovations Limited, suggests a period of active enforcement, which often precedes the filing of PTAB petitions. The patent is "Expired - Lifetime" as of January 16, 2023, according to Google Patents, which would typically preclude new IPR filings unless the petition was filed before the patent's expiration date or a CBM was available. However, considering the recent litigation filings (some in 2026), it is critical to investigate the exact nature of the asserted claims and the patent owner's infringement theories to determine the best defensive strategy.

Inventors

The named inventors for US Patent 6868682 are Ratnesh K. Sharma, Cullen E. Bash, and Chandrakant D. Patel. They initially assigned their interest in the patent application to HEWLETT-PACKARD COMPANY on the filing date, January 16, 2003 (Reel 014283/0337). This indicates they were likely employed by Hewlett-Packard Company at the time of filing. There is no information to suggest an unusual pattern of inventors departing the original assignee within 12 months of filing.

Original assignee

The original assignee listed on the issued patent is Hewlett Packard Development Co LP. Hewlett-Packard (HP) was a prominent operating company that manufactured and sold a wide array of computer hardware, software, and IT services. Their primary business involved providing technology products and solutions. In 2015, Hewlett-Packard Company split into two publicly traded companies: HP Inc. (focusing on personal systems and printers) and Hewlett Packard Enterprise (HPE, focusing on enterprise products and services). The patent remained within the HPE family after this split, specifically with Hewlett Packard Enterprise Development LP, both of which are currently operating companies.

Assignment timeline

• 2003-01-16 (executed) / recorded 2003-01-16 — Reel 014283/0337

  • Conveyance: ASSIGNMENT
  • Assignor: BASH, CULLEN E.; PATEL, CHANDRAKANT D.; SHARMA, RATNESH K.
  • Assignee: HEWLETT-PACKARD COMPANY
  • Correspondent: HEWLETT-PACKARD COMPANY, 3000 HANOVER STREET, PALO ALTO, CA 94304, US.
  • Context: Initial assignment from the inventors to their employer upon filing the patent application.

• 2003-06-12 (executed) / recorded 2003-06-18 — Reel 014389/0149

  • Conveyance: ASSIGNMENT
  • Assignor: HEWLETT-PACKARD COMPANY
  • Assignee: HEWLETT-PACKARD DEVELOPMENT COMPANY, L.P.
  • Correspondent: HEWLETT-PACKARD COMPANY, 3000 HANOVER STREET, PALO ALTO, CA 94304, US. This correspondent recurs in this chain.
  • Context: Internal corporate transfer within the Hewlett-Packard family of entities.

• 2003-09-24 (executed) / recorded 2003-09-30 — Reel 014876/0130

  • Conveyance: ASSIGNMENT
  • Assignor: HEWLETT-PACKARD COMPANY
  • Assignee: HEWLETT-PACKARD DEVELOPMENT COMPANY, L.P.
  • Correspondent: HEWLETT-PACKARD COMPANY, 3000 HANOVER STREET, PALO ALTO, CA 94304, US. This correspondent recurs in this chain.
  • Context: Another internal corporate transfer within the Hewlett-Packard family, possibly to refine ownership or include additional intellectual property.

• 2015-10-22 (executed) / recorded 2015-11-09 — Reel 034994/0638

  • Conveyance: ASSIGNMENT
  • Assignor: HEWLETT-PACKARD DEVELOPMENT COMPANY, L.P.
  • Assignee: HEWLETT PACKARD ENTERPRISE DEVELOPMENT LP
  • Correspondent: HEWLETT-PACKARD COMPANY, 3000 HANOVER STREET, PALO ALTO, CA 94304, US. This correspondent recurs in this chain.
  • Context: Internal corporate transfer following the separation of Hewlett-Packard Company into HP Inc. and Hewlett Packard Enterprise.

• 2021-04-12 (executed) / recorded 2021-05-06 — Reel 052441/0212

  • Conveyance: ASSIGNMENT
  • Assignor: HEWLETT-PACKARD DEVELOPMENT COMPANY, L.P.
  • Assignee: VALTRUS INNOVATIONS LIMITED
  • Correspondent: HEWLETT PACKARD ENTERPRISE DEVELOPMENT LP, 11445 Compaq Center Drive West, Houston, TX 77070, US. This correspondent is the legal department of the assignor's former corporate family.
  • Context: Transfer of the patent from the Hewlett Packard Enterprise family to Valtrus Innovations Limited.

Timeline diagram

timeline
    title Ownership of US 6868682
    2003 : Inventors to Hewlett-Packard Co
         : HP Co to HP Dev Co LP
         : HP Co to HP Dev Co LP
    2005 : Patent Issued
    2015 : HP Dev Co LP to HPE Dev LP
    2021 : HPE Dev LP to Valtrus Innovations

NPE / troll-pattern signals

1. Shell-entity transfer — Present. The patent was transferred to VALTRUS INNOVATIONS LIMITED on 2021-04-12 (executed) / 2021-05-06 (recorded) (Reel 052441/0212). Valtrus Innovations Limited is the plaintiff in numerous infringement lawsuits related to this patent, as indicated by the litigation summary, and is not known to be an operating company that produces products embodying the claims. The name "Innovations Limited" is also suggestive of a licensing or holding entity.
2. Known asserter in the chain — Present. Valtrus Innovations Limited is actively asserting this patent in district court litigation, as evidenced by the multiple pending cases listed in the litigation summary (e.g., Case 2: 1:26-cv-03958 filed June 21, 2026 [cite: https://portal.unifiedpatents.com/litigation/Illinois%20Northern%20District%20Court/case/1%3A26-cv-03958], Case 3: 2:26-cv-00287 filed March 10, 2026 [cite: https://portal.unifiedpatents.com/litigation/Texas%20Eastern%20District%20Court/case/2%3A26-cv-00287]).
3. Repeat correspondent across the chain — Present. The correspondent "HEWLETT-PACKARD COMPANY" (3000 HANOVER STREET, PALO ALTO, CA 94304, US) appears on four assignments within the HP corporate family: Reel 014283/0337 (2003-01-16), Reel 014389/0149 (2003-06-18), Reel 014876/0130 (2003-09-30), and Reel 034994/0638 (2015-11-09). The final assignment to Valtrus Innovations Limited (Reel 052441/0212, 2021-05-06) lists "HEWLETT PACKARD ENTERPRISE DEVELOPMENT LP" (11445 Compaq Center Drive West, Houston, TX 77070, US) as the correspondent, which is the legal department of the assignor or its corporate family. This consistent use of internal counsel for all transfers within and out of the HP ecosystem is noted.
4. Cascading transfers — Not present. The transfers within the HP family are spaced years apart, and the final transfer to Valtrus Innovations Limited is a single step.
5. Pre-litigation transfer — Present. The patent was transferred to Valtrus Innovations Limited (Reel 052441/0212) on April 12, 2021 (executed), with litigation by Valtrus beginning in September 2024 (e.g., 2:24-cv-00776 [cite: https://portal.unifiedpatents.com/litigation/Texas%20Eastern%20District%20Court/case/2%3A24-cv-00776]). While the gap (approximately 3 years) is longer than the typical 6-month window, the transfer to a known asserting entity clearly predates and enables the subsequent extensive litigation campaign, suggesting an intent to assert.
6. Bankruptcy fire-sale — Not present. The transfer was from an active operating company (Hewlett Packard Development Company, L.P.) and not part of bankruptcy proceedings.
7. Privateering — Unclear. While the patent originated from a major operating company (Hewlett-Packard/HPE) and was transferred to an asserting entity (Valtrus), there is no publicly available information (e.g., SEC filings, media reports) to definitively confirm a privateering arrangement where HPE is secretly funding or directing Valtrus's litigation.
8. Defensive aggregator (anti-NPE) — Not present. The chain terminates with Valtrus Innovations Limited, which is an asserting entity, not a defensive aggregator.

Verdict

NPE — high confidence

This verdict is based on several strong signals: the patent was transferred from an operating company (Hewlett Packard Development Company, L.P.) to Valtrus Innovations Limited (Reel 052441/0212) in 2021. Valtrus Innovations Limited is a known asserter, having initiated multiple infringement lawsuits concerning this patent beginning in late 2024 and continuing into 2026 [cite: https://portal.unifiedpatents.com/litigation/Texas%20Eastern%20District%20Court/case/2%3A24-cv-00776, https://portal.unifiedpatents.com/litigation/Illinois%20Northern%20District%20Court/case/1%3A26-cv-03958]. These actions, coupled with the name of the assignee, strongly indicate a licensing-focused entity rather than an operating company.

Verification of assignments can be performed at the USPTO Assignment Center: https://assignmentcenter.uspto.gov/

To identify the most relevant prior art for US Patent 6868682, I will examine the patent citations listed in the patent document itself. The USPTO's Patent Public Search database can be used to search for specific patents and their cited prior art.

"Prior art" refers to all publicly available information about existing inventions at the time a patent application is filed, and it is used to determine if an invention is novel and non-obvious (under 35 U.S.C. §§ 102 and 103, respectively).

I will now extract the patent citations directly from the text of US6868682 and analyze them.

Most Relevant Prior Art for US Patent 6868682

US Patent 6868682 references the following patent in its "CROSS-REFERENCE TO RELATED APPLICATIONS" section:

1. U.S. application Ser. No. 09/970,707
   • Full Citation: U.S. application Ser. No. 09/970,707, by Patel et al., filed, titled "SMART COOLING OF DATA CENTERS", and assigned to the assignee of the present invention.
   • Publication/Filing Date: The filing date for this application is not explicitly stated within US6868682, only that it is a co-pending application. However, as it is referenced, it would have been filed before US6868682's filing date of January 16, 2003.
   • Brief Description: This application is titled "SMART COOLING OF DATA CENTERS," indicating a similar subject matter to US6868682. Given that it is a co-pending application by the same inventors and assigned to the same entity, it likely describes foundational or related aspects of intelligent data center cooling systems.
   • Potential Anticipation (35 U.S.C. § 102): As a co-pending application by the same inventors and assignee, it is likely "secret prior art" under 35 U.S.C. § 102(e) (pre-AIA) if its effective filing date predates that of US6868682 and if it describes the claimed invention. If the subject matter claimed in US6868682 is fully disclosed in U.S. application Ser. No. 09/970,707, it could potentially anticipate any of the claims (Claims 1, 9, 10, 11, and 12) in US6868682. This would primarily depend on the overlap in the scope of claims between the two applications and whether the earlier application provides a complete disclosure of the later claimed invention.

Obviousness Analysis of US Patent 6868682 Under 35 U.S.C. § 103

The effective filing date for US Patent 6868682 is January 16, 2003. Obviousness under 35 U.S.C. § 103 requires determining whether the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art (PHOSITA). This analysis typically involves identifying a primary prior art reference, identifying a motivation to combine it with one or more secondary references (or common general knowledge), and demonstrating a reasonable expectation of success in achieving the claimed invention.

Limitation Due to Unavailable Prior Art:

US Patent 6868682 explicitly cross-references U.S. application Ser. No. 09/970,707, titled "SMART COOLING OF DATA CENTERS," by Patel et al., and assigned to the same assignee. This co-pending application is noted as a relevant piece of prior art in the "CROSS-REFERENCE TO RELATED APPLICATIONS" section. However, despite multiple attempts to retrieve the full publication details and text of U.S. application Ser. No. 09/970,707 via targeted searches on the USPTO Patent Center and Google Patents, the specific publication number and comprehensive content were not successfully accessed within the scope of this analysis. Therefore, a detailed obviousness analysis that directly compares the claims of US6868682 against the full disclosure of U.S. application Ser. No. 09/970,707 cannot be fully performed at this time.

General Obviousness Principles and Hypothetical Considerations:

Without the full text of the referenced U.S. application Ser. No. 09/970,707, any specific obviousness argument remains hypothetical. However, based on the descriptive title "SMART COOLING OF DATA CENTERS" and the fact that it shares inventors and assignee with US6868682, it is highly probable that it discloses core concepts related to intelligent data center cooling.

A PHOSITA in the field of data center cooling, at the time of the invention (before January 16, 2003), would have possessed knowledge of:

• Conventional data center cooling systems, often employing CRAC units operating at or near maximum capacity, irrespective of actual heat load.
• The inefficiency and high energy consumption associated with such conventional systems.
• The general concept of distributed sensing and control systems in various industrial applications.
• The use of hierarchy in control systems for managing complex environments.

If U.S. application Ser. No. 09/970,707 indeed describes a "smart cooling" system for data centers, it would likely have disclosed at least some elements of distributed sensing, localized control, and potentially hierarchical agent-based management for optimizing cooling.

Let's consider the independent claims of US6868682:

• Independent Claim 1 (Method): Describes a method for temperature control using a first agent to receive sensory data, process it, adjust cooling fluid delivery, and request a second agent from a hierarchy for redistribution if necessary.
• Independent Claim 9 (System): Describes a system with a cooling fluid delivery controller (first and second agents for rack and row control) and a cooling system controller (third agent for overall output) in a hierarchy.
• Independent Claim 10 (Method - Rack Specific): Focuses on a rack agent receiving sensor info, comparing to objectives, varying liquid cooling flow, and reporting to a row agent.
• Independent Claim 11 (Method - Row Specific): Focuses on a row agent receiving sensor info, comparing to objectives, varying liquid cooling flow in a row, and reporting to a CRAC agent or other row agents.
• Independent Claim 12 (Method - Cooling System/CRAC Specific): Focuses on a cooling system agent receiving sensor info, comparing to objectives, varying cooling liquid output, and reporting to other cooling system agents.

Hypothetical Obviousness Argument (assuming disclosure in U.S. application Ser. No. 09/970,707):

If U.S. application Ser. No. 09/970,707 (as "SMART COOLING OF DATA CENTERS") disclosed an intelligent cooling system for data centers that involved:

1. Distributed temperature sensing: Monitoring temperatures at various points within a data center, such as at individual racks or rows.
2. Localized control mechanisms: Adjusting cooling fluid delivery at a granular level (e.g., via variable vent tiles or localized fans).
3. A tiered or hierarchical control architecture: Where different levels of control (e.g., local, intermediate, global) coordinate to manage the cooling system.
4. Software agents: Employing autonomous or semi-autonomous software entities to implement these sensing and control functions.

A PHOSITA, at the time of the invention, would have found it obvious to combine these elements to achieve the claimed inventions in US6868682.

• Motivation to Combine: The motivation to combine these elements would stem from the recognized problems in conventional data center cooling: high energy consumption, inefficient cooling, and lack of responsiveness to localized heat loads. A "smart cooling" system, as implied by the title of the co-pending application, would inherently seek to address these issues. Implementing a hierarchical agent-based control system, as broadly described in the claims of US6868682, would be a logical and well-understood approach for managing the complexity and distributed nature of a data center cooling environment. The desire to optimize energy efficiency and cooling effectiveness would strongly motivate a PHOSITA to combine distributed sensors, local actuation (like adjustable vent tiles), and hierarchical software agents to create a more dynamic and responsive cooling system.
• Reasonable Expectation of Success: Given the state of the art in control systems and distributed computing at the time, a PHOSITA would have a reasonable expectation of success in implementing such an agent-based hierarchical control system. The foundational concepts of sensors, actuators, and software logic for decision-making were well-established. Adapting these to the specific problem of data center cooling, particularly within a tiered structure (rack, row, CRAC), would be considered a matter of engineering design choices rather than an inventive leap. The general concept of agents communicating and escalating requests for resources or assistance (e.g., a rack agent requesting more cooling from a row agent, or a row agent requesting more from a CRAC agent) is a standard design pattern in distributed control and artificial intelligence systems.

Conclusion on Obviousness (Pending Prior Art Review):

Without the specific content of U.S. application Ser. No. 09/970,707, it is not possible to conclusively state which claims of US6868682 would be rendered obvious by this reference alone or in combination with other prior art. However, if the referenced co-pending application broadly disclosed the use of "smart cooling" in data centers through distributed sensing and control, a PHOSITA would likely find it obvious to implement such a system using a hierarchical agent-based architecture to achieve the benefits of improved energy efficiency and localized temperature management as described in the independent claims of US6868682. The general principles of distributed control and hierarchical management were known in the art, and applying them to the specific problem of data center cooling, particularly in an effort to improve upon the "worst-case scenario" operation of conventional CRAC units, would be an expected step for a PHOSITA.

For US Patent 6868682, here's a breakdown of its term details:

Patent Term Adjustments (PTA)

Patent Term Adjustment (PTA) can extend the term of a U.S. patent to compensate for delays caused by the USPTO during the prosecution of a utility or plant patent application. This adjustment is added to the standard 20-year patent term from the earliest effective filing date. Delays can include the USPTO failing to:

• Issue an office action within 14 months of filing.
• Respond to a reply or appeal within four months.
• Act on an application within four months after a PTAB or federal court decision.
• Issue a patent within four months after issue fee payment.
• Issue a patent within 36 months from the filing date.

The USPTO automatically determines any PTA and provides notice of this determination no later than the patent's issuance date. Applicants have an opportunity to request reconsideration of the PTA determination.

To determine the exact PTA for US6868682, a direct search of the USPTO's Patent Center or Public PAIR database for this specific patent number would be required. This information is typically provided on the issued patent or in the patent's prosecution history. Without direct access to the USPTO's detailed records for US6868682, the specific number of PTA days cannot be precisely stated here.

Patent Term Extensions (PTE)

Patent Term Extensions (PTE) are distinct from PTAs and are granted under 35 U.S.C. § 156 to compensate for delays incurred in obtaining regulatory approval for a patented product (such as pharmaceuticals, medical devices, food additives, or color additives) from agencies like the FDA. The patent must claim the product, a method of using it, or a method of manufacturing it.

Key requirements for PTE include:

• The patent term must not have expired before the PTE application is submitted.
• The patent's term must not have been previously extended under 35 U.S.C. § 156(e)(1).
• The application for extension must be submitted by the patent owner or their agent within 60 days of regulatory agency approval.
• The product must have undergone a regulatory review period before commercial marketing or use.
• Only one patent can be extended for a regulatory review period for any product.

Given that US Patent 6868682 relates to an "Agent based control method and system for energy management" in data centers, it is highly unlikely to be eligible for a Patent Term Extension under 35 U.S.C. § 156, as its subject matter does not fall within the categories of products requiring pre-market regulatory review by agencies such as the FDA.

Continuation and Divisional Applications

The full patent text for US6868682 mentions a "CROSS-REFERENCE TO RELATED APPLICATIONS" section that refers to "U.S. application Ser. No. 09/970,707, by Patel et al., filed, titled 'SMART COOLING OF DATA CENTERS', and assigned to the assignee of the present invention." This indicates that US6868682 is not a continuation or divisional of an earlier application but rather claims priority to it. To definitively identify any continuation or divisional applications of US6868682, a direct search of the USPTO database for related family members would be necessary. Without this direct search, it is not possible to provide a comprehensive list of such applications.

Related Family Members

As mentioned, US Patent 6868682 references "U.S. application Ser. No. 09/970,707, by Patel et al., filed, titled 'SMART COOLING OF DATA CENTERS'." This is a related family member. To obtain a complete list of all related family members, including any continuations or divisionals that might stem from US6868682, a detailed family search within the USPTO database or commercial patent databases would be required. The Google Patents entry for US6868682 lists "US20040141542A1" as "Other versions," which is the publication of the application that matured into US6868682.

Projected Expiration Date

The term of a U.S. utility patent generally runs for 20 years from its earliest effective non-provisional filing date. For US Patent 6868682, the filing date is January 16, 2003.

Without any PTA, the base expiration date would be:
January 16, 2003 + 20 years = January 16, 2023.

The Google Patents listing for US6868682 explicitly states the "Anticipated expiration" as "2023-01-16" and the "Legal status" as "Expired - Lifetime". This aligns with the standard 20-year term from the filing date, suggesting that there were no significant Patent Term Adjustments that would extend its life beyond this date, or if there were, they were insufficient to prevent expiration on this date. As previously noted, this patent is not eligible for PTE. Therefore, based on the provided information and general patent term rules, US Patent 6868682 is confirmed to have expired on January 16, 2023.

Defensive Disclosure: Agent-Based Hierarchical Thermal Management Systems

This document describes various derivative works and technical disclosures building upon the core principles of US Patent 6868682, "Agent based control method and system for energy management." The objective of this defensive disclosure is to render future incremental improvements in hierarchical agent-based thermal management systems, particularly within data center environments and related fields, obvious or non-novel to a person having ordinary skill in the art. The focus is on expanding the scope of the original patent through material and component substitution, operational parameter expansion, cross-domain application, integration with emerging technologies, and analysis of inverse/failure modes.

---

Derivative 1: Liquid Immersion Cooling with Dielectric Fluids and Micro-Fluidic Actuation

Enabling Description:
This derivative implements the hierarchical agent-based control described in US6868682, replacing gaseous cooling fluids and macroscopic vent tiles with advanced liquid immersion cooling utilizing dielectric fluids. Each "subsystem" (e.g., individual server blades, GPUs, or processors) within a rack is fully submerged in a non-conductive, low-viscosity dielectric fluid such as engineered fluorocarbons (e.g., 3M Novec fluids) or synthetic mineral oils.

Rack Agents (analogous to the "first agent" in Claim 1 and "rack agent" in Claim 10) are embedded within each immersion tank. They receive real-time temperature data from localized thermistors (e.g., RTDs, thermocouples) strategically placed on critical components (CPUs, GPUs, memory modules). Instead of vent tiles, rack agents control micro-electromechanical systems (MEMS) based micro-pumps and electronically actuated micro-valves (e.g., piezoelectric or electro-osmotic) that regulate the flow rate of the dielectric fluid directly over individual heat-dissipating components or into localized cold plates/heat sinks within the tank. These micro-actuators enable highly granular and precise thermal management at the component level, varying the volumetric flow rate of the dielectric coolant. If a rack agent cannot maintain component temperatures within its predetermined range by manipulating local micro-valves and micro-pumps, it reports to its supervising Row Agent.

Row Agents (analogous to the "second agent" in Claim 1 and "row agent" in Claim 11) manage fluid distribution across an entire row of immersion-cooled racks. They receive aggregated thermal and flow data from subordinate rack agents. If a rack agent escalates a cooling demand, a row agent analyzes the overall thermal load and available cooling capacity within its row. It then adjusts the flow rate of the main dielectric fluid supply manifold to individual rack tanks using larger, electronically controlled proportional valves and variable-speed circulation pumps. This constitutes the redistribution of cooling fluid (as per Claim 1). Row agents can also balance cooling demand by dynamically re-routing excess cooled fluid from underutilized racks to those with higher thermal loads within the same row.

CRAC Agents (analogous to the "third agent" in Claim 9 and "cooling system agent" in Claim 12) oversee the entire primary dielectric fluid cooling loop. This includes managing large-scale, variable-capacity fluid chillers, heat exchangers (e.g., dry coolers or cooling towers), and bulk fluid filtration/purification systems. CRAC agents receive aggregated thermal loads and fluid pressure/flow data from row agents and plenum-equivalent sensors (e.g., main supply/return manifold temperature and pressure sensors). They adjust the chiller compressor speed, heat exchanger fan speeds, and main circulation pump speeds to maintain the overall dielectric fluid supply temperature and flow required by the data center's aggregated thermal load. If row agents cannot collectively meet their objectives, the CRAC agent increases the overall cooling liquid output, varying the amount of heat extraction from the dielectric fluid.

graph TD
    CRACA[CRAC Agent] -- Aggregated Thermal/Flow Data --> CRACONT[CRAC Controller (Chiller/HX/Pump)]
    CRACONT -- Adjusts Global Cooling Output (Dielectric Fluid) --> MainSupplyManifold[Main Dielectric Fluid Supply Manifold]

    MainSupplyManifold -- Distributes Fluid --> RowA1[Row Agent 1]
    MainSupplyManifold -- Distributes Fluid --> RowAN[Row Agent N]

    RowA1 -- Adjusts Row Manifold Flow --> RowManifold1[Row Supply Manifold 1]
    RowManifold1 -- Delivers Fluid to Rack Tanks --> RA1_Tank1[Rack Agent 1 / Immersion Tank 1]
    RA1_Tank1 -- Controls Micro-Pumps/Valves --> Comp1_1[Component 1.1 Temperature]
    RA1_Tank1 -- Controls Micro-Pumps/Valves --> Comp1_N[Component 1.N Temperature]
    Comp1_1 -- Thermistor Data --> RA1_Tank1
    Comp1_N -- Thermistor Data --> RA1_Tank1
    RA1_Tank1 -- Local Adjustment Failure/Report --> RowA1

    RowAN -- Adjusts Row Manifold Flow --> RowManifoldN[Row Supply Manifold N]
    RowManifoldN -- Delivers Fluid to Rack Tanks --> RAN_Tank1[Rack Agent N / Immersion Tank 1]
    RAN_Tank1 -- Controls Micro-Pumps/Valves --> CompN_1[Component N.1 Temperature]
    CompN_1 -- Thermistor Data --> RAN_Tank1
    RAN_Tank1 -- Local Adjustment Failure/Report --> RowAN

    RowA1 -- Aggregated Demand --> CRACA
    RowAN -- Aggregated Demand --> CRACA

    style CRACA fill:#f9f,stroke:#333,stroke-width:2px
    style RowA1 fill:#bbf,stroke:#333,stroke-width:2px
    style RowAN fill:#bbf,stroke:#333,stroke-width:2px
    style RA1_Tank1 fill:#dfd,stroke:#333,stroke-width:2px
    style RAN_Tank1 fill:#dfd,stroke:#333,stroke-width:2px

---

Derivative 2: Cryogenic Thermal Management for Superconducting Systems with Picosecond Response

Enabling Description:
This derivative extends the hierarchical agent-based control to ultra-low temperature, high-performance computing environments, specifically superconducting quantum computers or specialized research facilities requiring cryogenic cooling. The "cooling fluid" is liquid helium (LHe) or a multi-stage cryocooler system operating down to 4 Kelvin or millikelvin ranges.

Rack Agents are deployed at the individual cryostat level, each monitoring a cluster of superconducting qubits or ultra-cold experimental apparatus. Temperature sensors consist of ultra-sensitive cryogenic thermistors (e.g., Cernox, RuO2) integrated directly onto chip packages or sample holders. Rack agents (the "first agent") directly control localized micro-flow restrictors or miniature Joule-Thomson valves (e.g., MEMS-based phase change valves) that regulate the flow of compressed helium gas into a cryostat's expansion stages or direct injection of liquid helium. These actuators are designed for picosecond response times, adjusting cooling capacity based on minute temperature fluctuations indicative of quantum state decoherence or high-power operations. If a rack agent detects an inability to maintain its target temperature (ee.g., within µK range) or detects anomalous thermal spikes, it rapidly escalates a request to its supervising Row Agent.

Row Agents (the "second agent") manage arrays of interconnected cryostats or entire dilution refrigerator systems. They aggregate ultra-low temperature data from multiple rack agents and monitor the performance of intermediate cryocooler stages (e.g., pulse tube cryocoolers, Gifford-McMahon cryocoolers). Upon receiving an escalation, a row agent may redistribute cooling capacity by adjusting the duty cycle or cooling power of specific cryocoolers, or by dynamically routing colder helium streams from adjacent, less active cryostats via high-speed cryogenic manifold valves. They also coordinate pre-cooling stages and ensure efficient heat exchange between different temperature platforms.

CRAC Agents (the "third agent") oversee the entire cryogenic infrastructure, including bulk liquid helium storage dewars, helium liquefiers/recirculators, and high-pressure gas compressors. They receive aggregated cryogenic load demands and system health metrics from row agents, as well as critical parameters like helium reservoir levels, liquefaction rates, and compressor runtimes. CRAC agents adapt the overall helium liquefaction rate, adjust compressor staging, and manage heat rejection systems (ee.g., water cooling for compressor intercoolers) to ensure stable and sufficient supply of cryogenic fluid. In the event of an overall thermal excursion or a critical system failure (e.g., a helium leak detected by mass spectrometry), CRAC agents can initiate emergency procedures, such as rapid cool-down/warm-up protocols, or diversion of liquid helium from less critical systems to maintain integrity of high-priority quantum operations.

stateDiagram-v2
    direction LR
    Rack_Quibit_Temp: Qubit Temp (mK/µK)
    Rack_Qubit_Anomaly: Qubit Thermal Anomaly
    Rack_OK --> Rack_Mon: Temp within µK range
    Rack_Mon --> Rack_Hot: Temp > T_set + δT
    Rack_Mon --> Rack_Cold: Temp < T_set - δT

    Rack_Hot --> Rack_Adj_Flow: Increase LHe Flow/Valve (picoseconds)
    Rack_Cold --> Rack_Adj_Flow: Decrease LHe Flow/Valve (picoseconds)
    Rack_Adj_Flow --> Rack_Mon: Re-monitor

    Rack_Hot --> Row_Escalate_Hot: If Local Adj Fails
    Rack_Cold --> Row_Escalate_Cold: If Local Adj Fails
    Rack_Qubit_Anomaly --> Row_Escalate_Anomaly: Critical Anomaly Detected

    Row_Idle --> Row_Receive_Esc: Receive Rack Alert
    Row_Receive_Esc --> Row_Analyze: Analyze Row Load/Capacity
    Row_Analyze --> Row_Redistribute: Adjust Cryocooler/Manifold
    Row_Redistribute --> Row_Idle: Await Rack Update

    Row_Analyze --> CRAC_Escalate: If Row Capacity Exceeded
    Row_Receive_Esc --> CRAC_Escalate: Critical Row Anomaly

    CRAC_Ready --> CRAC_Receive_Esc: Receive Row Alert
    CRAC_Receive_Esc --> CRAC_Adjust_LHe: Adjust Liquefier/Compressor
    CRAC_Adjust_LHe --> CRAC_Ready: Await Row Update
    CRAC_Receive_Esc --> CRAC_Emergency: If Global Failure Detected
    CRAC_Emergency --> CRAC_Shutdown: Initiate Safe Shutdown/Diversion

    Rack_Mon -- Request -- Row_Idle
    Row_Idle -- Request -- CRAC_Ready

---

Derivative 3: Cross-Domain Application: Smart Environmental Control for Vertical Agriculture

Enabling Description:
This derivative applies the hierarchical agent-based control system to a vertical farming environment, optimizing environmental parameters for plant growth, specifically targeting temperature, humidity, and CO2 levels across multiple tiers of plant racks. The "cooling fluid" concept is generalized to environmental conditioning (air, nutrient solution temperature, and targeted airflow).

Rack Agents (the "first agent") are assigned to individual plant growth racks or hydroponic/aeroponic beds within a vertical farm. They receive sensory data from arrays of micro-environmental sensors, including infrared thermopiles for plant canopy temperature, relative humidity sensors, air temperature probes, and CO2 concentration sensors. Based on predefined plant-specific growth recipes and objectives, these rack agents directly control localized actuators: variable-speed micro-fans (e.g., EC fans for targeted airflow), ultrasonic misters for humidity control, precision CO2 injectors, and thermoelectric coolers/heaters (Peltier devices) for localized nutrient solution temperature regulation. If a rack agent cannot maintain the microclimate within optimal ranges (e.g., leaf temperature, root zone temperature, humidity setpoints), it escalates its needs to the supervising Row Agent.

Row Agents (the "second agent") oversee a vertical column or row of plant racks. They aggregate environmental data and demands from multiple rack agents. If a rack agent signals a persistent deviation, a row agent assesses the aggregated demand and redistributes environmental resources. This involves adjusting the main airflow plenums and louvers for a column of racks, varying the temperature of the primary nutrient solution supply, or initiating localized dehumidification/humidification cycles for the entire row. The row agent might also dynamically adjust grow light intensity or spectrum across the row to influence plant transpiration and heat load.

CRAC Agents (the "third agent" / "cooling system agent") manage the macro-environmental control for the entire vertical farm facility. They receive aggregated environmental demands and status reports from row agents. Their control parameters include the primary HVAC (Heating, Ventilation, Air Conditioning) system, large-scale CO2 generation/recirculation units, water chilling/heating plants for nutrient solution reservoirs, and overall air exchange systems. The CRAC agent optimizes energy consumption while ensuring all row-level environmental objectives are met. For example, if multiple row agents report high humidity, the CRAC agent increases the dehumidification capacity of the main HVAC unit. The goal is to provide overall environmental stability and resource efficiency across the entire agricultural operation.

graph TD
    CRACA[Farm Manager Agent (CRAC)] -- Aggregated Environmental Demands --> HVAC[HVAC Control]
    CRACA -- Aggregated Environmental Demands --> CO2Gen[CO2 Generator]
    CRACA -- Aggregated Environmental Demands --> WaterPlant[Nutrient Water Chiller/Heater]
    HVAC -- Adjusts Farm-Level Air Temp/Humidity --> FarmAir[Farm Air Environment]
    CO2Gen -- Adjusts Farm-Level CO2 --> FarmAir
    WaterPlant -- Adjusts Bulk Nutrient Temp --> BulkNutrient[Bulk Nutrient Solution]

    FarmAir -- Environmental Control --> RowA1[Row Agent 1]
    BulkNutrient -- Nutrient Supply --> RowA1
    FarmAir -- Environmental Control --> RowAN[Row Agent N]
    BulkNutrient -- Nutrient Supply --> RowAN

    RowA1 -- Microclimate Redistribution --> RowAirflow1[Row Airflow Plenum 1]
    RowA1 -- Microclimate Redistribution --> NutrientManifold1[Nutrient Manifold 1]
    RowAirflow1 -- Local Airflow --> RackA1_1[Rack Agent 1.1]
    NutrientManifold1 -- Local Nutrient Temp --> RackA1_1
    RackA1_1 -- Local Actuators (Fans, Misters, Peltier, CO2 Inj) --> PlantSensors1_1[Plant/Micro-Env Sensors 1.1]
    PlantSensors1_1 -- Sensor Data --> RackA1_1
    RackA1_1 -- Escalation --> RowA1

    RowAN -- Microclimate Redistribution --> RowAirflowN[Row Airflow Plenum N]
    RowAN -- Microclimate Redistribution --> NutrientManifoldN[Nutrient Manifold N]
    RowAirflowN -- Local Airflow --> RackAN_1[Rack Agent N.1]
    NutrientManifoldN -- Local Nutrient Temp --> RackAN_1
    RackAN_1 -- Local Actuators (Fans, Misters, Peltier, CO2 Inj) --> PlantSensorsN_1[Plant/Micro-Env Sensors N.1]
    PlantSensorsN_1 -- Sensor Data --> RackAN_1
    RackAN_1 -- Escalation --> RowAN

    RowA1 -- Aggregated Demands --> CRACA
    RowAN -- Aggregated Demands --> CRACA

    style CRACA fill:#f9f,stroke:#333,stroke-width:2px
    style RowA1 fill:#bbf,stroke:#333,stroke-width:2px
    style RowAN fill:#bbf,stroke:#333,stroke-width:2px
    style RackA1_1 fill:#dfd,stroke:#333,stroke-width:2px
    style RackAN_1 fill:#dfd,stroke:#333,stroke-width:2px

---

Derivative 4: AI-Driven Predictive Hierarchical Cooling with IoT and Blockchain Integration

Enabling Description:
This derivative enhances the agent-based hierarchical control system with advanced AI, IoT, and blockchain technologies for proactive and auditable thermal management in data centers.

Rack Agents (the "first agent" / "rack agent") are augmented with local edge AI models (e.g., small recurrent neural networks) running on embedded microcontrollers. These agents receive high-frequency, granular sensor data from a dense array of IoT sensors (e.g., LoRaWAN-enabled thermal arrays, localized airflow velocity sensors, power consumption meters) positioned on individual components within a rack. Instead of merely reacting to threshold breaches, these edge AI models perform real-time predictive analytics to anticipate localized hot spots or cooling deficiencies based on current workload patterns (communicated from server BMCs/hypervisors), historical thermal profiles, and predicted compute demands. This allows rack agents to proactively adjust local cooling fluid delivery (e.g., dynamic vent tiles or variable-speed rack fans) before a temperature threshold is crossed. All significant sensor readings, agent decisions, and actuator commands are cryptographically hashed and appended to a local, immutable blockchain ledger maintained by the rack agent. If predictive control or local adjustments are insufficient, the rack agent requests assistance from the row agent.

Row Agents (the "second agent" / "row agent") similarly incorporate larger, more complex AI models (e.g., deep learning models) that aggregate data from all subordinate rack agents. They perform zone-level predictive optimization, considering inter-rack airflow dynamics and potential cascading thermal effects. A row agent's AI can dynamically redistribute cooling fluid (e.g., adjusting row-level vent tiles, activating supplemental in-row cooling units) based on a global optimization function that balances thermal stability with energy efficiency for the entire row. The row agent also serves as a validator node for the rack agents' blockchain ledgers, securely aggregating and committing validated rack-level thermal events and actions to a higher-level row blockchain.

CRAC Agents (the "third agent" / "cooling system agent") host a central, powerful AI-driven optimization engine (e.g., a reinforcement learning agent) that continuously learns from the aggregated real-time IoT sensor data and blockchain-validated events from all row agents. This central AI predicts data center-wide thermal loads up to 24-48 hours in advance, considering external weather forecasts, electricity pricing, and anticipated global workload migrations (e.g., cloud bursting events). Based on these predictions, the CRAC agent proactively adjusts the overall cooling fluid output (e.g., chiller capacity, fan speeds, pump runtimes) to minimize energy consumption while guaranteeing desired thermal performance. The CRAC agent also maintains the master data center blockchain, consolidating all row-level blockchain data, providing a complete, auditable record of all thermal management decisions, sensor states, and energy usage. This blockchain can be used for compliance, energy credit trading, and verifying the provenance and maintenance history of cooling system components (supply chain verification).

sequenceDiagram
    participant IoT as IoT Sensors (Rack)
    participant RA as Rack Agent (Edge AI + Local Blockchain)
    participant RoA as Row Agent (Zone AI + Row Blockchain)
    participant CRACA as CRAC Agent (Central AI + DC Blockchain)
    participant Workload as Workload Scheduler/Hypervisor

    loop Real-time Monitoring & Prediction
        IoT->>RA: Stream High-Freq Temp/Flow/Power Data
        Workload->>RA: Current/Predicted Workload
        RA->>RA: Edge AI: Predict local hotspots
        RA->>RA: Adjust local cooling (vent/fan) proactively
        RA->>RA: Record decision/data to Local Blockchain
    end

    alt Local Adjustment Insufficient
        RA->>RoA: Escalation: Local Cooling Unsustainable
    end

    RoA->>RA: Request validated data (Local Blockchain)
    RA-->>RoA: Send Local Blockchain hashes/summaries
    RoA->>RoA: Zone AI: Aggregate data, analyze row dynamics
    RoA->>RoA: Optimize row cooling (redistribute fluid, in-row units)
    RoA->>RoA: Validate and commit data to Row Blockchain
    alt Row Adjustment Insufficient
        RoA->>CRACA: Escalation: Row Cooling Unsustainable
    end

    CRACA->>RoA: Request validated data (Row Blockchain)
    RoA-->>CRACA: Send Row Blockchain hashes/summaries
    CRACA->>CRACA: Central AI: Global predictive optimization (incl. external factors)
    CRACA->>CRACA: Adjust global cooling output (chillers, main fans)
    CRACA->>CRACA: Consolidate/Commit to Data Center Blockchain

    CRACA-->>CRACA: Audit/Compliance (from DC Blockchain)
    CRACA-->>CRACA: Energy Credit Trading (from DC Blockchain)
    CRACA-->>CRACA: Supply Chain Verification (from DC Blockchain)

---

Derivative 5: Adaptive Graceful Degradation and Low-Power Operations for Resilience

Enabling Description:
This derivative focuses on the "inverse" of optimal cooling: managing cooling during failures, low-power states, or emergency conditions. The hierarchical agent system is designed for adaptive graceful degradation and dynamic prioritization of cooling resources.

Rack Agents (the "first agent" / "rack agent") continuously monitor local component temperatures but also receive criticality ratings for the services running on their respective servers (e.g., "mission-critical," "essential," "non-essential," "idle"). In a graceful degradation scenario (triggered by a higher-level agent), a rack agent will dynamically increase its temperature setpoints for non-critical components, or even initiate selective power throttling/shutdown procedures for non-essential server workloads via integration with the server's Baseboard Management Controller (BMC) or hypervisor. This reduces local heat load. The rack agent will prioritize maintaining optimal temperatures for critical services by allocating the maximum available local cooling fluid (e.g., fully opening dynamic vent tiles or maximizing localized fan speeds) to those components, even if it means non-critical areas exceed their ideal temperature ranges or are temporarily starved of cooling. If local cooling cannot prevent critical components from overheating, the rack agent escalates.

Row Agents (the "second agent" / "row agent") detect aggregation of rack-level degradation or receive explicit commands from CRAC agents to enter a degradation mode. The row agent then dynamically re-prioritizes cooling fluid distribution within its row. For instance, if a CRAC agent signals reduced overall cooling capacity, the row agent might shift cooling fluid from "low-criticality" rack areas to "high-criticality" areas, even if it means some non-essential racks temporarily operate above their normal thermal limits or are effectively de-prioritized. The row agent can also instruct rack agents to collectively raise their low-criticality temperature thresholds to conserve overall cooling capacity. This involves a coordinated redistribution of remaining cooling resources to ensure survival of essential operations.

CRAC Agents (the "third agent" / "cooling system agent") are responsible for orchestrating the overall graceful degradation and emergency response. Upon detecting a major cooling system failure (e.g., compressor failure, significant power outage affecting cooling, or an instruction to enter a low-power mode for energy conservation), the CRAC agent initiates a predefined degradation profile. This involves:

1. Reduced Cooling Output: Systematically reducing chiller capacity, fan speeds, and pump runtimes to reflect available resources or target energy reduction.
2. Global Setpoint Adjustment: Communicating revised, elevated temperature setpoints to all row and rack agents, prioritizing critical infrastructure.
3. Workload Migration/Throttling: Interfacing with the data center's workload orchestration system (e.g., Kubernetes, OpenStack) to trigger migration of critical workloads to cooler racks (if available) or to initiate throttling/hibernation of non-critical workloads to reduce aggregate heat generation.
4. Emergency Ventilation: Potentially activating emergency ventilation systems (e.g., economizers, exhaust fans) in conjunction with reduced active cooling to passively manage heat in less critical areas.
5. Fault Isolation: Identifying and isolating failed cooling components or sections to prevent cascading failures, while routing remaining capacity around the fault.

The system ensures that even in degraded states, critical data center functions remain operational for as long as possible, foregoing optimal conditions for survival and essential service delivery.

stateDiagram-v2
    direction LR
    state "Normal Operation" as NormalOp {
        Rack_Normal: Rack Agents Maintain T_opt
        Row_Normal: Row Agents Balance T_opt
        CRAC_Normal: CRAC Agents Optimize P_cool
    }

    state "Graceful Degradation" as DegradedOp {
        Rack_Degrade: Prioritize Critical Workloads, T_set_non_crit_inc
        Row_Degrade: Redistribute to Critical Racks, T_set_row_inc
        CRAC_Degrade: Reduce P_cool, Global T_set_inc, Workload Throttling
    }

    state "Emergency Shutdown" as EmergencyStop {
        System_Halt: Rapid Safe Shutdown of Non-Critical, Maintain Critical Briefly
    }

    NormalOp --> DegradedOp: Cooling System Failure OR Low-Power Command
    DegradedOp --> EmergencyStop: Critical Component Overheat OR Prolonged Failure
    DegradedOp --> NormalOp: Cooling Restored OR Low-Power Mode Exited

    CRAC_Normal --> CRAC_Degrade: Global Failure / Low Power Event
    CRAC_Degrade --> Row_Degrade: Command: Adjust Row Priorities
    Row_Degrade --> Rack_Degrade: Command: Adjust Rack Priorities/Throttling

    Rack_Normal --> Rack_Degrade: Row/CRAC Command or Local Critical Overheat
    Rack_Degrade --> EmergencyStop: Critical Workload Thermal Limit Exceeded

    EmergencyStop --> System_Halt: System initiated shutdown

---

Combination Prior Art Scenarios

Here are three scenarios where the principles of US Patent 6868682, particularly its hierarchical agent-based control, are combined with existing open-source standards to establish prior art for integrated data center management.

1. Integration with OpenStack Nova/Neutron for Coordinated Compute and Cooling Resource Management:

   • Description: The hierarchical agent-based cooling system (rack, row, CRAC agents) is integrated as a service within an OpenStack cloud computing environment.
   • Mechanism:
     • Rack Agents expose APIs compatible with OpenStack Nova (compute service) and Neutron (networking service) to report real-time thermal conditions and receive workload context (e.g., VM/container ID, criticality, expected CPU/GPU utilization) from the hypervisor or guest OS via Nova.
     • Row Agents aggregate rack-level thermal data and can influence Nova's scheduler. For example, if a row agent anticipates a thermal bottleneck in a particular row, it can signal Nova to avoid placing new high-compute VMs in that row, or even trigger live migration of existing VMs to cooler rows (orchestrated by Nova/Neutron).
     • CRAC Agents act as a top-level resource provider for OpenStack's Resource Management. They can provide overall cooling capacity and energy consumption metrics to OpenStack for facility-level scheduling decisions or energy cost optimization. In turn, OpenStack can provide global workload forecasts to the CRAC agent for predictive cooling adjustments.
   • Open-Source Standard: OpenStack (Nova Compute, Neutron Networking, Heat Orchestration, Ceilometer Monitoring).
   • URL: https://www.openstack.org/

2. Agent Communication and Sensor Data Acquisition via MQTT/OPC UA:

   • Description: The hierarchical agents (rack, row, CRAC) and their associated sensors (temperature, pressure, flow) and actuators (vent tiles, fans, pumps, compressors) communicate using open-standard IoT/Industrial protocols.
   • Mechanism:
     • All sensors (thermistors, pressure transducers, flow meters) are equipped with microcontrollers that publish their readings as messages to a centralized or distributed MQTT Broker.
     • Rack Agents subscribe to MQTT topics relevant to their assigned rack's sensors and publish their control commands to specific actuator topics (e.g., a topic for a vent tile's aperture setting).
     • Row Agents subscribe to aggregated topics from multiple rack agents and publish commands to row-level actuators (e.g., manifold valves). They also act as OPC UA Servers, exposing aggregated thermal data and control points to higher-level enterprise systems or SCADA.
     • CRAC Agents subscribe to row-level MQTT topics and can act as OPC UA Clients to gather data from row agents, while also exposing their global control parameters (e.g., chiller setpoints, overall fan speed) as an OPC UA Server for facility-wide management. This standardization ensures interoperability and reduces proprietary lock-in for sensors and actuators.
   • Open-Source Standard: MQTT (Message Queuing Telemetry Transport), OPC UA (Open Platform Communications Unified Architecture).
   • URLs:
     • MQTT: https://mqtt.org/
     • OPC UA: https://opcfoundation.org/

3. Applying Agent-Based Control to Building Management Systems (BMS) with BACnet/Modbus:

   • Description: The hierarchical agent system manages a data center's cooling infrastructure by integrating with existing building automation and control networks using standard industrial communication protocols.
   • Mechanism:
     • Existing Computer Room Air Conditioning (CRAC) units, Chillers, and facility-level fans/pumps are typically controlled via a Building Management System (BMS) using protocols like BACnet or Modbus.
     • CRAC Agents in this scenario do not directly control the physical CRAC components but rather interact with the BMS's BACnet or Modbus interfaces (acting as a client) to request adjustments to cooling output (e.g., changing supply air temperature setpoints, fan speed commands). The CRAC agent's objectives are translated into BACnet/Modbus write commands to the BMS.
     • Row Agents and Rack Agents still directly control localized vent tiles or in-row cooling units. However, they can also report their aggregated thermal demand and cooling fluid flow requirements to the CRAC agent, which then communicates these high-level requests to the BMS. For instance, a persistent high thermal load reported by a row agent might lead the CRAC agent to request a lower supply air temperature from the BMS.
     • The BMS acts as an intermediary, translating the agent's high-level cooling requests into specific control signals for the legacy CRAC units and other building infrastructure. This enables the smart, agent-based control to operate within existing data center facilities without requiring a complete overhaul of the legacy cooling equipment.
   • Open-Source Standard: BACnet (Building Automation and Control Network), Modbus.
   • URLs:
     • BACnet: https://www.bacnet.org/
     • Modbus: https://modbus.org/