DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
This office action is in response to amendment/reconsideration filed on 08/13/2026, the amendment/reconsideration has been considered. Claims 1, 2, 5-6, 13-14 and 20 have been amended. Claims 1-20 are pending for examination as cited below.
Response to Arguments
Applicant's arguments filed on 08/13/2026 have been fully considered but they are not persuasive. In remarks applicant argues in substance that:
(a) Applicant argues that “Illikkal” merely classifies workloads as CPU intensive, memory intensive, network intensive, storage intensive, etc. and therefore does not disclose the claimed resource requirements.
Examiner respectfully disagree because Illikkal teaches collecting detailed operating information from node computing resources, grouped computing resources, applications, and fIllikkaleworks. Such operating information includes processor utilization, memory utilization, network I/O characteristics, storage bandwidth, storage latencies, processor affinity, throttling information, and workload execution characteristics. Illikkal further teaches building workload models using this collected operational data and generating workload profiles based on historical monitoring information. Specifically, Illikkal taches collecting operational information including processor, memory, storage and network resources utilization information and building analytical and simulation models representing workload behavior. Illikkal further teaches classifying workloads based on such collected workload characteristics and workload profiles e.g., [0019-0013] disclose collection of operating information from CPUs, memory, storage devices, network devices, virtual machines, applications and software. In para [0027-0029] disclose workload classification based on monitored information and creating of workload profiles such as CPU intensive, memory intensive, network intensive and storage intensive profiles.
The amended claim does not require direct measurement under a physically contention free environment. Rather, it requires determining resource requirements associated with the workload. Illikkal’s workload models are generated from monitored workload execution characteristics and resource consumption behavior and therefore, reasonably represent workload demand characteristics for processor, memory, storage and network resources. One of ordinary skilled in the art would have understood that such workload modeling inherently requires determining resource consumption characteristics attributable to the workload itself in order to build and apply the predictive models taught by Illikkal. Accordingly, Illikkal teaches or at least suggests determining workload resource requirements associated with multiple computing resources.
(b) Applicant argues that “Chen” only teaches generic provisioning of resources and does not teach resource availability for each compute node comprising an amount of each resource available to a workload on that node.
Examiner respectfully disagree because Chen teaches determining resource availabilities across physical machines, virtual machines, servers, and provisionable resources. Chen teaches evaluating resource availability by identifying idle CPUs, idle memory, available applications, available software licenses, available virtual machine capacity, network resources and storage resources for workload scheduling decision. Chen further analyzes resource availabilities of physical machines, virtual machines and servers to determine whether resources satisfying workload requirements are available or capable of being provisioned. See, for example, discussions of resources 142, 147, 155, 157 and 165, and the analysis preformed during workload scheduling.
Specifically, Chen states that scheduling progIllikkal 200 “analyzes resource availabilities of physical machines 140 and 145, and servers 150 and 160” and determines available CPUs, memory, software, licenses, and other resources associated with individual resources and machines. See discussions corresponding to step 202.
While Chen may not use the phrase “microarchitecture resource availability” it clearly teaches determining, for individual computing resources and machines, the amount of available resources that may be allocated to a workload. It would have been obvious to use the availability information generated by Chen as inputs to the predictive scheduling fIllikkalework or Illikkal and Nagpal because both references are directed toward improving workload placement and resource allocation decision.
(c) Applicant Argues that Nag’s model operates on workload schedules and system configuration information and does not utilize workload resource requirements and node specific resource availability representations as required by the claims.
Examiner respectfully disagree because, Nag teaches a resource performance predictive model that receives workload information, workload schedules, system configuration information, resource schedule data, configuration data, workload performance correlations and configuration performance correlations to predict resource performance characteristics and facilitate resource scheduling. See fig.2-5 and the associated description.
Nag specifically teaches training a resource performance predictive model based on workload inputs and resource responses; generating workload performance correlations; adapting the predictive model to differing computing environments and using the adapted resource performance predictive model to perform resource scheduling.
Additionally, Nag explains that resource schedule data and configuration data are inputs to the predictive model, and that the model predicts performance characteristics responsive to workloads consuming compute and storage resources. Nagpal further teaches that the predictive model is used to facilitate workload placement and resource scheduling decisions. See abstract, fig.1a, 1b and corresponding description.
The claimed “workload signature modle” reads not Nag’s resource performance predictive model. The exact wording differs, but the functionality is substantially the same because both receive workload related representations and system resource/configurations representations and produce performance predictions.
(d) Applicant yet again argues that none of the cited references explicitly disclose workload performance for each candidate compute node and selecting one compute node based on the respective predicted performance.
Examiner respectfully disagree because the argument is not persuasive when teachings are considered collectively under 35 U.S.C. 103 rejection.
Illikkal teaches: predicting behavior of a data center under different operation scenarios; generating predicted performance characteristics; providing those predictions to resource managers, schedulers, and orchestrators and using the predictions to facilitate resource allocation and scheduling decision see paragraphs [0012], [0025], [0031-0035].
Chen teaches: evaluating resource availability across multiple physical machines and virtual machines; determining which resources satisfy workload requirements, and scheduling workloads among candidate resources and machines.
Nag teaches: generating resource performance predictions based on workload and configuration input; adapting the predications to specific target environment and configurations, and using the predictions to perform resource scheduling, see Abstract, fig.1a, fig.2.
A person of ordinary skill would have found it obvious to apply Nag’s predictive model separately to the multiple candidate resource or nodes identified by Chen using workload information supplied by Illikkal in order to compare expected performance among those candidate nodes before scheduling the workload. There merely represents the predictable use of prior art predictive scheduling techniques to evaluate alternative placement candidates.
Thus, the combined references teach resource requirements (Illikkal); determining node specific resource availability (Chen); applying a predictive workload/resource performance model using workload and configuration/resource information (Nag); generating predicted performance values for candidate placement options; and selecting a compute node based on the predicted performance results (Illikkal + Chen + Nag).
Accordingly, applicant’s arguments have been fully considered but are not persuasive. The combination of Illikkal, Chen, and Nag collectively teaches all limitations of amended independent claims 1, 13 and 20. The amendment do not patentably distinguish the claimed subject matter from the cited prior art because the references collectively disclose workload characterization, node/resource availability determination, predictive performance modeling using workload and resource information, and selection of resources for workload placement based on predicted performance outcomes. Therefore, the rejection of claims 1, 13 and 20 and claims dependent therefrom, under 35 U.S.C. 103 is maintained.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if 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 to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-4 and 6-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Illikkaleshkumar et al. (Pub. No.: US 2017/0286252 A1), hereinafter “Illikkal” in view of Chen et al. (Patent No.: US 9396028 B2), hereinafter “Chen” and further in view of Nagpal et al. (Patent No.: US 10691491 B2), hereinafter “Nag”.
As to claim 1. Illikkal discloses, an apparatus (Illikkal, Abstract and fig.1) comprising:
processing circuitry to perform operations comprising, upon receipt of a primary workload to be scheduled on a data center (Illikkal, fig.1, [0012], cloud scout manager monitors and drives model-based scheduling):
determining resource requirements associated with the primary workload, the resource requirements comprising, for each of one or more microarchitecture resources, an amount of the resource consumed by the primary workload in the absence of resource contention (Illikkal, [0023], and [0027-0029], Model logic and workload profiling build workload profiles and models that represent per-workload behavior);
a performance associated with running the primary workload on the compute node (Illikkal, fig.6, [0023-0025], model(s) built from monitored operating information are used to predict data-center behavior under scenarios and to facilitate allocation/scheduling); and
selecting one of the compute nodes for placement of the primary workload based at least in part on the predicted performance associated with the compute nodes (Illikkal, fig.4, [0025], indicate logic provides predicted behavior to job scheduler/resource manager to facilitate placement decisions); and
a network interface to transmit the primary workload to the selected one of the compute nodes (Illikkal, fig.8, [0007], System architecture includes communications interface and orchestration to enact placement).
Illikkal however is silent to disclose explicitly, determining resource availabilities for a cluster of compute nodes, the resource availability for each compute node comprising an amount of each of the one or more microarchitecture resources that is available to the primary workload on the compute node (Chen, col.2 lines 5-13, identifying a workload in the set of pending workloads that is scheduled to utilize hypothetic resources, wherein hypothetic resources are idle computer resources that are currently not available, but can be made available to execute workloads through provisioning actions; holding the identified workload from dispatch to hypothetic resources for a holding period, wherein the holding period is a customizable duration of time.)
Therefore, before the effective filing date of the instant application it would have been obvious to one of the ordinary skilled in the art to incorporate the teachings of “Chen” into those of “Illikkal” to provide a computer-implemented method for workload scheduling and resource provisioning. In accordance with the present invention, the computer implemented method includes the steps of scheduling a set of pending workloads for execution on computer resources in a computing environment to enhance the process of workload scheduling.
Illikkal and Chen however are silent to disclose explicitly, operating a workload signature model on representations of the determined resource requirements of the primary workload and the resource availability of the compute node to predict.
Nag however discloses explicitly, operating, for each of the compute nodes, a workload signature model (Nag, col.2 lines 46-53) on representations of the determined resource requirements (Nag, col.7, lines 20-25) and the resource availabilities (Nag, fig.4, col.4, lines 34-41) to predict, for each compute node, a performance associated with running the primary workload on the compute node (Nag, col.4, lines 42-50 and col.4, lines 50-57 and col.5, lines 18-22, dynamic scheduling and workload placement adjustment and fig.6, a cluster having multiple nodes 652.).
Therefore, before the effective filing date of the instant application it would have been obvious to one of the ordinary skilled in the art to incorporate the teachings of “Nag” into those of “Illikkal and Chen” to provide a Systems for distributed resource system management. A first computing system operates in a first computing environment. A predictive model is trained in the first computing environment to form a trained resource performance predictive model that comprises a set of trained model paIllikkaleters to capture at least computing and storage IO paIllikkaleters that are responsive to execution of one or more workloads that consume computing and storage resources in the first computing environment. When the trained resource performance predictive model is deployed to a second computing environment, various computing system configuration differences, and/or workload differences and/or other differences between the first computing environment and the second computing environment are detected and measured.
As to claim 2. The combined system of Illikkal, Chen and Nag discloses the invention substantially including, wherein determining the resource requirements comprises (Illikkal, [0023], operating information monitored by node monitoring 151-1, infrastructure 151-2, fIllikkalework monitoring 151-3, software monitoring 151-4 or application monitoring 151-5 may be collected and at least temporarily stored in database 152. In some examples, model logic 153 may build one or more model(s) 153-1 using the collected operating information stored in database 152). causing the primary workload to be temporarily run alone on one of the compute nodes and receiving measurements of the resource requirements from that compute node (Chen fig.2, block 200-214, discloses, scheduling, provisioning, and dispatch based on measured/provisioned resources and supports the overall scheduling context in which measured workload runs would be used to determine requirements. Also see Nag, fig.3, steps 308-312, which describes running workloads in a controlled environment to obtain measured resource responses that become the workload signature used for scheduling.).
As to claim 3. The combined system of Illikkal, Chen and Nag discloses the invention substantially including, wherein determining the resource availabilities comprises receiving, from the compute nodes, measurements of the resource availabilities in the presence of background workloads on the compute nodes (Nag, fig.3 and 5A, 502-514, col.13, lines 24-34, the detected environment change event invokes a modification of the then-current model paIllikkaleters (e.g., the training model paIllikkaleters 134 or derivatives thereof) to dynamically adapt the resource performance predictive model to the target environment (step 512). In certain embodiments, the model paIllikkaleters are modified based on the environment differences associated with the environment change event. A learning phase of the resource performance predictive model 132.sub.2 can commence using the adapted model paIllikkaleters (e.g., dynamically modified model paIllikkaleters 136) (step 514).).
As to claim 4. The combined system of Illikkal, Chen and Nag discloses the invention substantially including, wherein the resource requirements and resource availabilities comprise amounts for multiple microarchitecture resources (Nag, fig.1A, Fig.2 “predictive model”, fig.3, “training flow blocks 306-312” and fig.4 “training data table”.).
As to claim 6. The combined system of Illikkal, Chen and Nag discloses the invention substantially as applied above including, wherein selecting one of the compute nodes for placement of the primary workload comprises comparing the predicted performances associated with running the primary workload on the compute nodes against a target value of a performance metric, and wherein the node selected for placement has a measured resource availability sufficient to meet the target value when executing the primary workload (Nag, Abstract, fig.1A step 110, fig.3 training/deploy/use; Illikkal, fig.6, [0024-0025], indicate logic provides predictive behavior and “what-if” insights to facilitate allocation against SLA/QoS requirements. The passages teach computing predicted performance per node and comparing/using those prediction relative to target SLO/QoS values. Further, Nag, fig.5A, collects local measurement so the scheduler has measured availability data).
As to claim 7. The combined system of Illikkal, Chen and Nag discloses the invention substantially as applied above including, wherein the target value of the performance metric is a hardware service level objective (SLO) derived from a performance guarantee associated with the primary workload pursuant to a service level agreement (SLA) (Illikkal, fig.1, fig.6 block 602-606, [0023-0025], requirements 154 may include predefined rules or polices associated with service level agreement (SLA) 154-1, quality of service (QoS) 154-2, or reliability, availability and serviceability (RAS) 154-3 requirements.).
As to claim 8. The combined system of Illikkal, Chen and Nag discloses the invention substantially as applied above including, wherein selecting one of the compute nodes for placement of the primary workload is further based on a cluster-level optimization policy (Illikkal, fig.1, 4, [0024-0025], requirements 154 may include predefined rules or polices associated with service level agreement (SLA) 154-1, quality of service (QoS) 154-2, or reliability, availability and serviceability. Nag teaches fig.1a , step-110, fig.2, fig.5a e.g. model deployment/adaptation and use of scheduling).
As to claim 9. The combined system of Illikkal, Chen and Nag discloses the invention substantially as applied above including, wherein the workload signature model comprises a machine-learned model (Illikkal, [0023]).
As to claim 10. The combined system of Illikkal, Chen and Nag discloses the invention substantially as applied above including, wherein the machine-learned model is based on training data (Nag, Abstract, fig.3, fig.4) comprising, for each of a plurality of collocation scenarios between primary and background workloads, associated measured resource availability and resource requirement vectors correlated with measured performance values (Nag, fig.4, training data tables, fig.3, blocks 306-312, shows training data and model data including node configuration, resource metrics and performance metrics. Nag further shows in fig.4, training data and “workload-performance correlation that explicitly links observed resource responses to performance outcomes. Illikkal, fig.6, [0023-0025] discloses predict/indicate logic uses model outputs (predicted behavior) and compares them to SLA/QoS targets.).
As to claim 11. The combined system of Illikkal, Chen and Nag discloses the invention substantially as applied above including, wherein the processing circuitry is at least in part configured by instructions stored in one or more computer-readable media to perform the operations (Illikkal, [0045]).
As to claim 12. The combined system of Illikkal, Chen and Nag discloses the invention substantially as applied above including, wherein the processing circuitry comprises one or more hardware accelerators to implement at least part of the operations (Illikkal, [0010] and apparatus 500 / circuitry 520.).
As to claim 13 is rejected for same rationale as applied to claim 1 above.
As to claim 14 is rejected for same rationale as applied to claim 2 above.
As to claim 15 is rejected for same rationale as applied to claim 3 above.
As to claim 16 is rejected for same rationale as applied to claim 4 above.
As to claim 17 is rejected for same rationale as applied to claim 6 above.
As to claim 18 is rejected for same rationale as applied to claim 8 above.
As to claim 19 is rejected for same rationale as applied to claim 10 above.
As to claim 20 is rejected for same rationale as applied to claim 1 above.
Claim(s) 5 is/are rejected under 35 U.S.C. 103 as being unpatentable over “Illikkal, Chen and Nag” as applied to parent claims above in view of Madduri et al. (Pub. No.: US 20190196827 A1), hereinafter “Madd”.
As to claim 5. The combined system of Illikkal, Chen and Nag discloses the invention substantially including, the multiple microarchitecture resources (Illikkal, fig.6, [0023], collecting per-node operating metrics).
Illikkal, Chen and Nag however are silent to disclose explicitly, last-level chance (LLC) and memory bandwidth.
Madd discloses a similar concept in the same field of endeavor including, last-level chance (LLC) and memory bandwidth (Madd, [0096], The set of shared cache units 606 may include one or more mid-level caches, such as level 2 (L2), level 3 (L3), level 4 (L4), or other levels of cache, a last level cache (LLC), and/or combinations thereof.).
Therefore, before the effective filing date of the instant application it would have been obvious to one of the ordinary skilled in the art to incorporate the teachings of “Madd” into those of “Illikkal, Chen and Nag” to provide method for performing signed multiplication of packed signed doublewords and accumulation with a signed quadword. For example, one embodiment of a processor comprises: a first source register to store a first plurality of packed signed doubleword data elements; a second source register to store a second plurality of packed signed doubleword data elements; a third source register to store a plurality of packed signed quadword data elements.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Please see the attached PTO-892.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to TAUQIR HUSSAIN whose telephone number is (571)270-1247. The examiner can normally be reached M-F 7:00 - 8:00 with IFP.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Vivek Srivastava can be reached on 571 272-7304. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/Tauqir Hussain/Primary Examiner, Art Unit 2446
