Microsoft Corporation v. ParTec AG

Prosecution History analysis

Examiner Search analysis

Invalidity Framework analysis

Petition

Petition summary

## 1. Case Identification
- Case Number: [IPR2025-00318](https://ai-lab.exparte.com/case/ptab/IPR2025-00318)
- Patent #: [11,537,442](https://ai-lab.exparte.com/patent/11537442)
- Filed: December 20, 2024
- Petitioner(s): [Microsoft Corporation](https://ai-lab.exparte.com/party/microsoft-corp)
- Patent Owner(s): [Partec Cluster Competence Center GmbH](https://ai-lab.exparte.com/party/partec-ag)
- Challenged Claims: 1-10

## 2. Patent Overview
- Title: Application Runtime Determined Dynamical Allocation Of Heterogeneous Compute Resources
- Brief Description: The ’442 patent relates to a method for operating a cluster computing system composed of heterogeneous processors, specifically "computation nodes" (e.g., multi-core CPUs) and "booster nodes" (e.g., accelerator-type processors). The core of the invention is a method to dynamically re-distribute computational sub-tasks among these different nodes between processing iterations, using information gathered from a preceding iteration to improve the efficiency of a subsequent iteration.

## 3. Grounds for Unpatentability
#### Ground 1: Claims 1-10 are obvious over Lippert in view of Budenske.
- **Prior Art Relied Upon:** [Lippert](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1004) (Application # 2013/0282787) and [Budenske](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1005) (Journal of Supercomputing, Oct. 1998).
- **Core Argument for this Ground:**
    - **Prior Art Mapping:** Petitioner argued that the combination of [Lippert](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1004) and [Budenske](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1005) taught all elements of the challenged claims. [Lippert](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1004) was cited for its disclosure of a heterogeneous cluster computing architecture with "computation nodes" and "boosters," along with a resource manager for assigning tasks between them. [Budenske](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1005), published nearly two decades before the ’442 patent’s priority date, was argued to supply the allegedly novel concept: a method for executing *iterative* tasks where sub-tasks are dynamically re-mapped between iterations based on performance data ("dynamic parameters") gathered from the preceding iteration, with the explicit goal of reducing execution time. Petitioner contended that combining [Lippert's](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1004) architecture with [Budenske's](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1005) iterative optimization method rendered independent claim 1 and its dependent claims obvious.
    - **Motivation to Combine:** A Person of Ordinary Skill in the Art (POSITA) would combine [Lippert's](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1004) flexible hardware architecture with [Budenske's](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1005) iterative re-mapping method to achieve the well-known design goal of improving the performance of iterative applications. Petitioner noted that [Budenske](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1005) expressly motivated implementing its method in flexible, heterogeneous computing systems like [Lippert's](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1004) to "fully utilize the architectural flexibility." Both references followed the same principle of outsourcing computationally intensive tasks to specialized processors, making the combination a natural and logical step for a POSITA.
    - **Expectation of Success:** Petitioner asserted a POSITA would have a high expectation of success. The combination involved implementing known software functions on well-understood hardware architectures. Further, other available prior art publications (e.g., [Kwok](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1007), [Theys](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1008)) discussed [Budenske's](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1005) approach, confirming its feasibility and effectiveness, which would have assured a POSITA that the combination was not only possible but beneficial.

#### Ground 2: Claims 2-5 and 8-10 are obvious over Lippert in view of Budenske and Kambatla.
- **Prior Art Relied Upon:** [Lippert](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1004) (Application # 2013/0282787), [Budenske](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1005) (Journal of Supercomputing, Oct. 1998), and [Kambatla](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1006) (Application # 2018/0074855).
- **Core Argument for this Ground:**
    - **Prior Art Mapping:** This ground built upon Ground 1, adding [Kambatla](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1006) to provide more explicit teachings for specific dependent claims relating to system management components. [Kambatla](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1006) taught a resource management framework for distributed clusters (akin to Apache YARN) that expressly disclosed an "application manager," a "resource manager," and per-node "daemons" for monitoring and reporting resource usage and status. Petitioner argued that [Kambatla’s](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1006) disclosure of an explicit `application manager` confirmed the obviousness of claim 2. Similarly, [Kambatla’s](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1006) teachings on a central resource manager and per-node daemons that generate loading information were used to show the obviousness of claims 3, 5, and 8.
    - **Motivation to Combine:** Petitioner argued that a POSITA seeking to implement the system of [Lippert](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1004) and [Budenske](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1005) would naturally turn to a known, practical framework like that taught by [Kambatla](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1006) for implementation details. [Kambatla](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1006) provided a blueprint for managing resources, handling multiple applications, and facilitating the inter-node communication required for the dynamic re-mapping taught by [Budenske](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1005). Incorporating [Kambatla's](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1006) teachings would be a logical step to efficiently manage the cluster and better leverage its full processing capacity.
    - **Key Aspects:** This ground aimed to reinforce the obviousness arguments of Ground 1 by demonstrating that the specific system management components recited in the dependent claims were not only logical extensions but were features of well-known, contemporary resource management architectures.

## 4. Arguments Regarding Discretionary Denial
- Petitioner argued that discretionary denial under *Fintiv* would be improper. It asserted that the petition was filed promptly after receiving infringement contentions in a parallel district court case. Petitioner contended the court case is in its early stages, with no significant discovery having occurred. It further argued that the scheduled trial date is likely to be delayed beyond the IPR's Final Written Decision (FWD) deadline, minimizing any potential overlap and inefficiency. Finally, Petitioner stressed the compelling merits of the petition, which relies on prior art ([Budenske](https://ai-lab.exparte.com/case/ptab/IPR2025-00318/doc/1005)) that was not considered by the Examiner and which allegedly teaches the core novelty of the challenged claims.

## 5. Relief Requested
- Petitioner requests institution of an inter partes review (IPR) and cancellation of claims 1-10 of Patent [11,537,442](https://ai-lab.exparte.com/patent/11537442) as unpatentable on the grounds specified.

Microsoft targets a heterogeneous computing patent, arguing all 10 claims are obvious over Lippert & Budenske prior art teaching iterative task re-mapping.