DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
This Office Action is in response to Applicant’s Amendment and Remarks filed on 27 June 2025.
Claims 1-7, 9-12, 14-15, 17-18 and 21-24 are pending for examination. Claims 8, 13, 16 and 19-20 were cancelled. Claims 21-24 were newly added.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-7, 9-12, 14-15, 17-18 and 21-24 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
Claim 1 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1, Statutory Category: Yes, the claim 1 is an apparatus that including different components for performing different steps and therefore falls in the statutory category of a machine.
Step 2A- Prong 1: Judicial Exception Recited: Yes, the claim recites: “modify resource allocation to perform second operations of a workload based on performance of first operations of the workload; modify resources allocated to perform the second operations”. As drafted, the claim as a whole recites a server that including different components for performing different steps that could be performed in the human mind, but for the recitation of generic computing components. The human mind can easily judging/modifying/planning/scheduling/changing the allocated resources based on previous determined performance of the operation of the workload. Therefore, but for the recitation of generic computing components, these steps may be a Mental Processes that can be performed in the human mind (including an observation, evaluation, judgment, opinion).
Therefore, yes, the claims do recite judicial exceptions.
Step 2A- Prong 2: Integrated into a practical Application: No, this judicial exception is not integrated into a practical application. In particular, the claim recites an additional limitations that “a network interface device comprising a direct memory access (DMA) circuitry, network interface, host interface, and a processor, wherein: the processor is to issue an instruction to a workload orchestrator to”, “based on the instruction, the workload orchestrator is to” and “the resources comprise disaggregated compute and memory resources accessible by Ethernet packets, and the resources comprise at least one of: a graphics processing unit (GPU), a general purpose GPU (GPGPU), a field programmable gate array (FPGA),accelerator, or central processing unit (CPU)” are directed to Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(f)) and an attempt to generally link the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05(h))). The combination of these additional elements is no more than mere instructions to apply the exception using a generic computer component (MPEP 2106.05(f)). Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application because they not impose any meaningful limits on practicing the abstract idea. Therefore, the claim is directed to the abstract idea.
Step 2B: Claim provides an Inventive Concept: No. The additional elements “a network interface device comprising a direct memory access (DMA) circuitry, network interface, host interface, and a processor, wherein: the processor is to issue an instruction to a workload orchestrator to”, “based on the instruction, the workload orchestrator is to” and “the resources comprise disaggregated compute and memory resources accessible by Ethernet packets, and the resources comprise at least one of: a graphics processing unit (GPU), a general purpose GPU (GPGPU), a field programmable gate array (FPGA),accelerator, or central processing unit (CPU)” are directed to Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(f)) and an attempt to generally link the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05(h))). These additional elements and combination of the elements does not amount to significant more than the exception itself or provide an inventive concept in Step 2B.
For these reasons, there is no inventive concept in the claim, and thus the claim is ineligible.
Independent claims 10 and 21 are rejected for the same reason as claim 1 above. In addition, independent claim 10 further recites “a plurality of compute resources”, “one or more non-transitory machine-readable storage media comprising a plurality of instructions stored thereon that, in response to being executed by the processor, cause the processor to”. These additional elements are directed to generic computing components/Functions (MPEP § 2106.05(b) merely applying the abstract idea (MPEP § 2106.05(f)).
With respect to the dependent claim 2, the claim elaborates that wherein the network interface device and workload orchestrator to use machine learning to profile workloads, specify a redundancy level for operations of the workload, and determine a configuration for resources allocated to perform the second operations(“use machine learning to profile workloads, specify a redundancy level…and determine” as being treated as part of abstract idea and is analogous to Mental processes, such that concept can be performed in the human mind).
With respect to the dependent claim 3, the claim elaborates that wherein the network interface device is to issue the instruction for allocation of resources based on workload training data, the workload, and resource and network utilization (“allocation of resources based on workload training data” as being treated as part of abstract idea and is analogous to Mental processes, such that concept can be performed in the human mind).
With respect to the dependent claim 4, the claim elaborates that wherein the network interface device is to use machine learning to generate workload training data (“use machine learning to generate workload training data” are directed to Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(f)) and an attempt to generally link the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05(h))).
With respect to the dependent claim 5, the claim elaborates that wherein second network interface device to use a configuration for operations stored in the workload training data to determine resource allocation for a workload phase (“use a configuration for operations stored in the workload training data to determine resource allocation” as being treated as part of abstract idea and is analogous to Mental processes, such that concept can be performed in the human mind).
With respect to the dependent claim 6, the claim elaborates that wherein the workload orchestrator specify a redundancy level for the operations based on a security level of the operations (“specify a redundancy level…based on a security level” as being treated as part of abstract idea and is analogous to Mental processes, such that concept can be performed in the human mind).
With respect to the dependent claim 7, the claim elaborates that wherein the workload orchestrator comprises a workload manager, the workload manager to compare a current operation with workload training data to allocate resources and identify a redundancy level for the current operation(“wherein the workload orchestrator comprising a workload manager” are directed to Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(f)) and an attempt to generally link the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)). In addition, “compare a current operation with workload training data” and “identify a redundancy level” as being treated as part of abstract idea and is analogous to Mental processes, such that concept can be performed in the human mind).
With respect to the dependent claim 9, the claim elaborates that wherein the network interface device is a member of a pool of network interface devices, the pool of network interface devices to communicate with the workload orchestrator as a single entity (“wherein the network interface device is a member of a pool of network interface devices, the pool of network interface devices to communicate with the workload orchestrator as a single entity” are directed to Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(f)) and an attempt to generally link the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)).
Dependent claims 11-12, 14-15 and 18 recite the similar features as applied to claims 2-3, 5-6 and 9 respectively above, therefore they are also rejected under the same rationale.
With respect to the dependent claim 17, the claim elaborates that wherein if a first compute node determines that additional resources are required, a second network interface device in a second compute node is to submit a resource allocation request to allocate more resources (“if a first compute node determines that additional resources are required, a second network interface device in a second compute node is to submit a resource allocation request to allocate” are directed to Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(f)) and an attempt to generally link the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05(h))).
Dependent claims 22-23 recite the same features as applied to claims 2 and 3 above, therefore they are also rejected under the same rationale.
With respect to the dependent claim 24, the claim elaborates that the resource allocation comprises a redundancy level, wherein the redundancy level is to specify a number of nodes that perform the operations (“wherein the redundancy level is to specify a number of nodes that perform the operations” are directed to Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(f)) and an attempt to generally link the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05(h))).
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
Claims 1-7, 9-12, 14-15, 17-18 and 21-24 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement.
The claims 1, 10 and 21 contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for pre-AIA the inventor(s), at the time the application was filed, had possession of the claimed invention. Because the specification fails to disclose that the network interface device comprising a direct memory access (DMA) circuitry and a processor of the network interface device is to issue an instruction to a workload orchestrator to modify resource allocation,.
More specifically, in claim 1 (lines 4-6), claim 10 (lines 2-8), claim 21 (lines 1-7), it recites “a network interface device comprising a direct memory access (DMA) circuitry, network interface, host interface, and a processor, wherein: the processor is to issue an instruction to a workload orchestrator to modify resource allocation”. Pages 13-14 of specification discloses “the NIC 532 may be embodied as part of a system-on-a-chip (SoC) that includes one or more processors…a network interface includes a network interface controller or a network interface card. In some examples, a network interface can include one or more of a network interface controller (NIC) 532, a host fabric interface (HFI), a host bus adapter (HBA), network interface connected to a bus or connection”. Page 23 of specification discloses “The IPU 1004 characterizes utilization of resources for currently running jobs and can submit resource allocation requests to the workload orchestrator 1110 based on its needs” and Page 21 of specification discloses “IPUs can be integrated with smart NICs and storage or memory (for example, on a same die, system on chip (SoC), or connected dies)”. The specification does NOT have support for the network interface device comprises a direct memory access (DMA) circuitry. And the specification fails to discloses how to using the processor of the network interface device to issue an instruction to a workload orchestrator to modify resource allocation. The specification at the best, it discloses that the IPU characterizes utilization of resources for currently running jobs and can submit resource allocation requests to the workload orchestrator”, the IPU issue the request to the workload orchestrator is NOT the same as the processor of the network interface device issue the request to the workload orchestrator. The specification does not indicating that the IPU is the processor within the network interface device (i.e.., the specification only indicates that IPUs can be integrated with smart NICs and storage or memory on a same die, system on chip (SoC), it does not necessary means that the IPU is the processor of the network interface device. They are two separated components integrated within the SOC). Therefore, the specification fails to disclose that the network interface device comprising a direct memory access (DMA) circuitry and a processor of the network interface device is to issue an instruction to a workload orchestrator to modify resource allocation.
Claims 2-7, 9, 11-12, 14-15, 17-18 and 22-24, they are depend on claims 1, 10 and 21 and do not overcome the deficiencies thereof, therefore they are rejected for the same reason as claims 1, 10 and 21 above.
Claim Rejections - 35 USC § 103
The following is a quotation of pre-AIA 35 U.S.C. 103(a) which forms the basis for all obviousness rejections set forth in this Office action:
(a) A patent may not be obtained though the invention is not identically disclosed or described as set forth in section 102, if the differences between the subject matter sought to be patented and the prior art are such that the subject matter as a whole would have been obvious at the time the invention was made to a person having ordinary skill in the art to which said subject matter pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 9-10, 18 and 21 are rejected under 35 U.S.C. 103 as being unpatentable over Schmidt et al. (US Pub. 2019/0213099 A1) in view of Dormitzer (US Patent. 2019/0065401 A1).
Schmidt was cited in the previous Office Action.
As per claim 1, Schmidt teaches the invention substantially as claimed including An apparatus comprising (Schmidt, Fig. 6):
a network interface device (Schmidt, [0052] lines 4-11, The compute resources of the cluster controller machine 651 include one or more processors each having one or more processor cores…The network resources of the cluster controller machine 651 can include, e.g., a physical network interface controller), wherein:
the processor is to issue an instruction to a workload orchestrator to perform resource allocation to perform second operations of a workload based on performance of first operations of the workload (Schmidt, Fig. 6, 651 cluster controller machine (as workload orchestrator), 654 ML-based resource prediction, 658 resource allocation, 664 reinforcement learning, 612A application scheduler (worked with CPU 602A) managed by 658 resource allocation of cluster controller machine; Fig. 4, 430 feed acquired training data into model as input, 440 use acquired training data to generate predications of resource usage; [0042] lines 1-5, combination of data pertaining to the load on hardware resources, e.g. CPU, memory, and I/O resources, and program behavior, as evidenced by the system calls made by an application over a time period t; [0045] lines 5-6, generate resource usage predictions for each of one or more applications during a future time period; [0051] lines 22-26, Each application behavior monitor and each application scheduler is a set of processor executable instructions stored at a processor readable memory, e.g. one of memories 604A and 604N, and configured to be executed by a processor, e.g. one of CPUs 602A and 602N; [0048] lines 2-12, the process acquired training data. The training data can be acquired, e.g., by recording for multiple time periods, data pertaining to load over time on physical resources and data pertaining to system events—as described above in connection with FIG. 2. At 420, the process initializes model parameters θ. Thereafter, at 430 the process feeds the acquired training data that corresponds to one or more first time periods (as performance of first operations of the workload) into the model as input, and at 440, the model uses the acquired training data to generate predictions of resource usage for one or more second time periods (i.e. performs a forward pass) (as second operations of the workload));
based on the instruction, the workload orchestrator is to perform resources allocation to perform the second operations (Schmidt, Fig. 5, 530, 540; [0050] lines 6-11, At 530, the process uses the model for predicting future resource usage to predict resource usage for one or more application during a future time period. At 540, the process allocates resources within the computing system according to the resource usage predicted at 530 and so as to achieve a certain performance target, e.g. minimize average waiting time, provided by a value function having parameters Q. Allocating resources at 540 can be performed on a per-application basis as well as at other levels of granularity), and
the resources comprise at least one of: a graphics processing unit (GPU), a general purpose GPU (GPGPU), a field programmable gate array (FPGA),accelerator, or central processing unit (CPU) (Schmidt, [0025] lines 11-18, Application-specific resource usage information can be of the same type as the system-wide resource usage information but on a per-application basis. For example, application-specific resource usage information can include current CPU and memory resources used by individual applications and an average over a sliding window of CPU and memory resources used by individual applications).
Schmidt fails to specifically teach a network interface device comprising a direct memory access (DMA) circuitry, network interface, host interface, and a processor, when performing the resource allocation, it is to modify resource allocation/modify resource allocated and the resources comprise disaggregated compute and memory resources accessible by Ethernet packets,
However, Dormitzer teaches a network interface device comprising a direct memory access (DMA) circuitry, network interface, host interface, and a processor (Dormitzer, Fig. 14, 830 comm circuit (as network interface device), 832 NIC; Fig. 16, 1652 NIC, 1654 memory access logic unit (as direct memory access (DMA circuitry); [0080] lines 21-26, The memory access logic unit 1654 may be embodied as any device or circuitry (e.g., an application specific integrated circuit (ASIC), a specialized processor, etc.) capable of enabling the NIC 1652 to perform direct memory access operations; [0052] lines 10-18, the NIC 832 may be embodied as part of a system-on-a-chip (SoC) that includes one or more processors…In some embodiments, the NIC 832 may include a local processor (not shown) and/or a local memory (not shown) that are both local to the NIC 832. In such embodiments, the local processor of the NIC 832 may be capable of performing one or more of the functions of the processors 820; [0088] lines 1-10, The communication circuitry 1702, in the illustrative embodiment, includes a network interface controller (NIC) 1652, which may also be referred to as a host fabric interface (HFI) (as host interface). The NIC 1652 may be embodied as one or more add-in-boards, daughter cards, network interface cards, controller chips, chipsets, or other devices that may be used by the accelerator sled 1640 to connect with another compute device (e.g., the compute sled 1630, the orchestrator server 1620, etc.); [0094] lines 1-17, network communicator 1820, which may be embodied as hardware, firmware, software, virtualized hardware, emulated architecture, and/or a combination thereof as discussed above, is configured to facilitate inbound and outbound network communications (e.g., network traffic, network packets, network flows, etc.) to and from the accelerator sled 1640, respectively. To do so, the network communicator 1820 is configured to receive and process data packets from one system or computing device (e.g., the compute sled 1630, the orchestrator server 1620, etc.)…at least a portion of the functionality of the network communicator 1820 may be performed by the communication circuitry 1702, and, in the illustrative embodiment, by the NIC 1652),
When performing the resource allocation, it is to modify resource allocation/modify resource allocated) (Dormitzer, [0075] lines 28-50, the orchestrator server 1520 may selectively allocate and/or deallocate physical resources 620 from the sleds 400 and/or add or remove one or more sleds 400 from the managed node 1570 as a function of quality of service (QoS) targets (e.g., performance targets associated with a throughput, latency, instructions per second, etc.) associated with a service level agreement for the workload (e.g., the application 1532). In doing so, the orchestrator server 1520 may receive telemetry data indicative of performance conditions (e.g., throughput, latency, instructions per second, etc.) in each sled 400 of the managed node 1570 and compare the telemetry data to the quality of service targets to determine whether the quality of service targets are being satisfied. If the so, the orchestrator server 1520 may additionally determine whether one or more physical resources may be deallocated from the managed node 1570 while still satisfying the QoS targets, thereby freeing up those physical resources for use in another managed node) and
the resources comprise disaggregated compute and memory resources accessible by Ethernet packets (Dormitzer, Fig. 14, 1400 memory sled; [0027] data center 100 in which disaggregated resources may cooperatively execute one or more workloads (e.g., applications on behalf of customers) includes multiple pods 110, 120, 130, 140, each of which includes one or more rows of racks… By disaggregating resources to sleds comprised predominantly of a single type of resource (e.g., compute sleds comprising primarily compute resources, memory sleds containing primarily memory resources), and selectively allocating and deallocating the disaggregated resources to form a managed node assigned to execute a workload, the data center 100 provides more efficient resource usage over typical data centers comprised of hyperconverged servers containing compute, memory, storage and perhaps additional resources). As such, the data center 100 may provide greater performance (e.g., throughput, operations per second, latency, etc.) than a typical data center that has the same number of resources; [0028] lines 18-20, the sleds in the pod 110 may still maintain data communication with the remainder of the data center 100 (e.g., sleds of other pods) through the other switch 250, 260. Furthermore, in the illustrative embodiment, the switches 150, 250, 260 may be embodied as dual-mode optical switches, capable of routing both Ethernet protocol communications (as Ethernet packets) carrying Internet Protocol (IP) packets).
It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to have combined the teaching of Schmidt with Dormitzer because Dormitzer’s teaching of providing disaggregated resources with a network interface controller for resource modification would have provided Schmidt’s system with the advantage and capability to allow the system to selectively allocating and deallocating the disaggregated resources to form a managed node assigned to execute a workload in order to provide more efficient resource usage over typical data centers comprised of hyperconverged servers (see Dormitzer, [0028]).
As per claim 9, Schmidt and Dormitzer teach the invention according to claim 1 above. Dormitzer further teaches wherein the network interface device is a member of a pool of network interface devices, the pool of network interface devices to communicate with the workload orchestrator as a single entity (Dormitzer, Fig. 14, 830 Comm circuit, 832 NIC; Fig. 15, 1540, 1550, 1560 (each including network interface devices (as pool of network interface devices); 1520 Orchestrator server; [0071] lines 1-15, in some embodiments, the sled 400 may be embodied as a memory sled 1400. The storage sled 1400 is optimized, or otherwise configured, to provide other sleds 400 (e.g., compute sleds 800, accelerator sleds 1000, etc.) with access to a pool of memory (e.g., in two or more sets 1430, 1432 of memory devices 720) local to the memory sled 1200. For example, during operation, a compute sled 800 or an accelerator sled 1000 may remotely write to and/or read from one or more of the memory sets 1430, 1432 of the memory sled 1200 using a logical address space that maps to physical addresses in the memory sets 1430, 1432. The memory sled 1400 includes various components similar to components of the sled 400 and/or the compute sled 800, which have been identified in FIG. 14 using the same reference numbers. The description of such components provided above in regard to FIGS. 6, 7, and 8 apply to the corresponding components of the memory sled 1400 and is not repeated herein for clarity of the description of the memory sled 1400).
As per claim 10, it is a system claim of claim 1 above. Therefore, it is rejected for the same reason as claim 1 above. In addition, Schmidt further teaches one or more non-transitory machine-readable storage media comprising a plurality of instructions stored thereon that, in response to being executed, cause the system to (Schmidt, Claim 15, A non-transitory computer readable medium having stored thereon computer executable instructions for performing a method for monitoring resources in a computing system having system information, the method comprising).
As per claim 18, it is a system claim of claim 9 above. Therefore, it is rejected for the same reason as claim 9 above.
As per claim 21, it is a method claim of claim 1 above. Therefore, it is rejected for the same reason as claim 1 above.
Claims 2 and 22 are rejected under 35 U.S.C. 103 as being unpatentable over Schmidt and Dormitzer, as applied to claims 1 and 21 respectively above, and further in view of Lacombe et al. (US Patent. 6,055,647).
Lacombe was cited in the previous Office Action.
As per claim 2, Schmidt and Dormitzer teach the invention according to claim 1 above. Schmidt further teaches wherein the network interface device and workload orchestrator to use machine learning to profile workloads and determine a configuration for resources allocated to perform the second operations (Schmidt, [0052] lines 4-11, The compute resources of the cluster controller machine 651 include one or more processors each having one or more processor cores…The network resources of the cluster controller machine 651 can include, e.g., a physical network interface controller [0034] lines 1-18, Information pertaining to the behavior of applications, as represented by system or library calls…pertaining to the application's behavior can be represented by a counting vector that counts the number of occurrences of each call during a pre-determined time period. Such pre-determined time period can be linked to time periods during which process resource usage metrics are measured; [0051] lines 22-26, Each application behavior monitor and each application scheduler is a set of processor executable instructions stored at a processor readable memory, e.g. one of memories 604A and 604N, and configured to be executed by a processor, e.g. one of CPUs 602A and 602N); [0045] lines 1-6, the fixed-dimension vector of system load measurements and the fixed-dimension vector representation of the system event data set are provided as input to a machine learning model that has previously been trained to generate resource usage predictions for each of one or more applications during a future time period; [0040] lines 11-15, To provide a simple model for the use case of resource usage prediction, an LSTM can be trained with collected and preprocessed data that includes past usage statistics and past system call data. The learning of system call embeddings can be performed as a preprocessing step or within an end-to-end architecture; [0050] lines 6-11, At 530, the process uses the model for predicting future resource usage to predict resource usage for one or more application during a future time period. At 540, the process allocates resources within the computing system according to the resource usage predicted at 530 and so as to achieve a certain performance target, e.g. minimize average waiting time, provided by a value function having parameters Q. Allocating resources at 540 can be performed on a per-application basis as well as at other levels of granularity]; please note: modifying resources was taught by Dormitzer).
Schmidt and Dormitzer fail to specifically teach specify a redundancy level for operations of the workload.
However, Lacombe teaches specify a redundancy level for operations of the workload (Lacombe, Col 6, lines 57-64, if additional load devices 12 are added to the computer system 10 or if the power consumption of the load components 12 increases during periods of heavy operation (as operations of the workload), redeterminations of the level of power supply redundancy are made, thereby to provide a dynamic indication of the actual level of power supply redundancy; Col 8, lines 40-41, determination of the level of power supply redundancy to be made).
It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to have combined the teaching of Schmidt and Dormitzer with Lacombe because Lacombe’s teaching of specifying the level of power redundancy for the workload phases would have provided Schmidt and Dormitzer’s system with the advantage and capability to allow the system to preventing the system failure due to the heavy operation in order to improving the system reliability and performance. (see Lacombe, Col 3, lines 21-24, “significantly enhance the reliability and serviceability aspects of a high-end server system”).
As per claim 22, it is a method claim of claim 2 above. Therefore, it is rejected for the same reason as claim 2 above.
Claims 3-4 and 7 are rejected under 35 U.S.C. 103 as being unpatentable over Schmidt, Dormitzer and Lacombe, as applied to claim 2 above, and further in view of SUBRAMANIAN et al. (US Pub. 2020/0104184 A1).
SUBRAMANIAN was cited in the previous Office Action.
As per claim 3, Schmidt, Dormitzer and Lacombe teach the invention according to claim 2 above. Schmidt, Dormitzer and Lacombe fail to specifically teach wherein the network interface device is to issue the instruction for allocation of resources based on workload training data, the workload, and resource and network utilization.
However, SUBRAMANIAN teaches wherein the network interface device is to issue the instruction for allocation of resources based on workload training data, the workload, and resource and network utilization (SUBRAMANIAN, Fig. 5, 520, 522 resource determination, suggested resource configuration to pod manager; [0044] lines 10-14, Resource determination module 522 can use an AI or ML model dedicated to determining and suggesting resource allocations for particular categories of applications; [0025] lines 8-15, workload table 224 keeps track of previously performed workloads and their characteristics such as one or more of: boundedness (e.g., utilization of one or more of: processor, memory, network (as network utilization), storage, or cache), applied resource allocations, telemetry data, or workload performance characteristic(s) (as workload). AI model 222 runs its inference model (as network interface device) and produces a predicted best or recommended resource configuration for that workload; also see [0021] lines 6-20, telemetry data that can be captured using counters or performance monitoring events related to: processor or core usage statistics, input/output statistics for devices and partitions…telemetry data such as but not limited to outputs from Top-down Micro-Architecture Method (TMAM), execution of the Unix system activity reporter (sar) command, Emon command monitoring tool that can profile application and system performance. However, additional information can be collected such as outputs from a variety of monitoring tools including but not limited to output from use of the Linux perf command, Intel PMU toolkit, Iostat, VTune Amplifier, or monCli or other Intel Benchmark Install and Test Tool (Intel® BITT) Tools. Other telemetry data can be monitored such as, but not limited to, power usage, inter-process communications, and so forth (as training data)).
It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to have combined the teaching of Schmidt, Dormitzer and Lacombe with SUBRAMANIAN because SUBRAMANIAN’s teaching of recommending the resource allocation/configuration based on system, workload performance and resource utilization would have provided Schmidt, Dormitzer and Lacombe’s system with the advantage and capability to allow the system to recommending the resource allocation based on different data which improving the determination accuracy and system performance.
As per claim 4, Schmidt, Dormitzer, Lacombe and SUBRAMANIAN teach the invention according to claim 3 above. In addition, SUBRAMANIAN further teaches wherein the network interface device is to use machine learning to generate workload training data (SUBRAMANIAN, Fig. 5, 520, Fig. 3, 302 AI model; [0025] lines 15-20, AI model 222 does not have to be trained prior to providing resource configuration recommendations as it will be configured to continuously learn on-the-fly using a reinforcement learning scheme; [0026] lines 1-5, AI model 222 keeps evolving and learning as it progressively receives different inputs. Reinforcement learning can be a goal-oriented scheme that learns how to attain a complex objective (goal) or maximize along a particular dimension over many steps (as use machine learning to generate the workload training data (i.e., reinforcement learning scheme)).
As per claim 7, Schmidt, Dormitzer, Lacombe and SUBRAMANIAN teach the invention according to claim 4 above. Schmidt further teaches wherein the workload orchestrator comprises a workload manager, the workload manager to compare a current operation with workload training data to allocate resources (Schmidt, Fig. 6, 660 and 662 compare to prediction); [0052] lines 4-7, The compute resources of the cluster controller machine 651 include one or more processors each having one or more processor cores; [0052] lines 22-30, Each of information collector 652, machine-learning based resource predictor 654, anomaly detector 656, resource allocator 658, resource usage monitor 660, prediction assessor 662, and reinforcement learning module 664 is a set of processor executable instructions stored at a processor readable memory, e.g. the memory resources of the cluster controller machine 651, and configured to be executed by a processor (as whole as workload manager); [0048] lines 13-16, compares the predictions of resource usage for the one or more second time periods (as training data, see Fig. 6, 664 to 654, reinforcement learning back to ML-based resource predication) with actual resource usage measured during the one or more second time periods; also see Fig. 5, 553 compare predications of resource usage with actual resource usage, 554 update model; and back to 515 to 540 allocation resources (as to allocate resources based on comparing and updating)).
In addition, Lacombe teaches identify a redundancy level for the current operation (Lacombe, Col 6, lines 57-64, if additional load devices 12 are added to the computer system 10 or if the power consumption of the load components 12 increases during periods (as including current workload phase) of heavy operation redeterminations of the level of power supply redundancy are made, thereby to provide a dynamic indication of the actual level of power supply redundancy; Col 8, lines 40-41, determination of the level of power supply redundancy to be made; also see Col 1, lines 14-15, the determination of the power supply redundancy level is made dynamically (as including current operation).
It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to have combined the teaching of Schmidt, Dormitzer and SUBRAMANIAN with Lacombe because Lacombe’s teaching of specifying the level of power redundancy for the workload phases would have provided Schmidt, Dormitzer and SUBRAMANIAN’s system with the advantage and capability to allow the system to preventing the system failure due to the heavy operation in order to improving the system reliability and performance. (see Lacombe, Col 3, lines 21-24, “significantly enhance the reliability and serviceability aspects of a high-end server system”).
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Schmidt, Dormitzer, Lacombe and SUBRAMANIAN, as applied to claim 3 above, and further in view of Radovanovic et al. (US Pub. 2021/0273858 A1).
Radovanovic was cited in the previous Office Action.
As per claim 5, Schmidt, Dormitzer, Lacombe and SUBRAMANIAN teach the invention according to claim 3 above. Dormitzer teaches second network interface device (Dormitzer, Fig. 8, 800, 830 comm circuit, 832 NIC; Fig. 15m different sled each including NIC (as including second network interface device).
Schmidt, Dormitzer, Lacombe and SUBRAMANIAN fail to specifically teach wherein second network interface device to use a configuration for operations stored in the workload training data to determine resource allocation for a workload phase.
However, Radovanovic teaches wherein second network interface device to use a configuration for operations stored in the workload training data to determine resource allocation for a workload phase (Radovanovic, Fig. 1, 102 computing device, 112 processor, 120 Machine-learned model, 162 training data; [0005] line 4, one or more computing devices; [0090] lines 7-11, a training computing system associated with the network computing system can receive historical training data (e.g., training data that can include information associated with the past states of the plurality of nodes over a plurality of time intervals) (as configuration stored in the workload training data); [0117] lines 17-25, The one or more operations performed by the computing device 102 can also include determining, based at least in part on the network data and a machine-learned model, the resource availability and the resource usage for at least the portion of the plurality of nodes at a time interval subsequent to the initial time interval. Furthermore, the one or more operations performed by the computing device 102 can include generating, based at least in part on the network data, one or more predictions for the portion of the plurality of nodes of the plurality of nodes (as determining resource allocation a workload phase (i.e., subsequence interval)); [0198] lines 11-19, based on data associated with the one or more predictions (e.g., a predicted future resource usage) the computing device 102 can generate one or more control signals that can be used to activate one or more devices and/or systems associated with providing and/or generating the resource within the network. For example, based on one or more predictions that the demand for a resource (e.g., electrical power) will decrease in one hour, the network computing system can send one or more signals to begin a reduction in the amount of electrical power).
It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to have combined the teaching of Schmidt, Dormitzer, Lacombe and SUBRAMANIAN with Radovanovic because Radovanovic’s teaching of using the training data regarding to the each workload phases (i.e., resource utilization of previous time intervals) for determine the resource allocation for subsequence time interval would have provided Schmidt, Dormitzer, Lacombe and SUBRAMANIAN’s system with the advantage and capability to allow the system to improve the performance of network operation by more effectively determining the state of the network and thereby providing information about the network that can be used to avoid situations in which portions of the network are overtaxed (see Radovanovic [0111] “improve the performance of network operation by more effectively determining the state of the network and thereby providing information about the network that can be used to avoid situations in which portions of the network are overtaxed”).
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Schmidt, Dormitzer and Lacombe, as applied to claim 2 above, and further in view of Warfield et al. (US Pub. 2014/0025770 A1).
Warfield was cited in the previous Office Action.
As per claim 6, Schmidt, Dormitzer and Lacombe teach the invention according to claim 2 above. Schmidt teaches the workload orchestrator (Schmidt, Fig. 6, 651 cluster controller machine (as workload orchestrator), see arrows communicating with 601A and 610N cluster machine); [0020] lines 1-3, a variety of resource allocators, e.g. process schedulers in operating systems and orchestrators in computing clusters).
In addition, Lacombe teaches specify a redundancy level for the operations based on a load level of the operations (Lacombe, Col 6, lines 57-64, if additional load devices 12 are added to the computer system 10 or if the power consumption of the load components 12 increases during periods of heavy operation (as each workload phase), redeterminations of the level of power supply redundancy are made, thereby to provide a dynamic indication of the actual level of power supply redundancy; Col 8, lines 40-41, determination of the level of power supply redundancy to be made).
Schmidt, Dormitzer and Lacombe fail to specifically teach when specify a redundancy level, it is based on a security level.
However, Warfield teaches when specify a redundancy level, it is based on a security level (Warfield, [0164] lines 8-12, The configured number may be a predetermined number set by an administrator or a static or dynamic value that is a function of the required level of security or redundancy that may be required to meet one or more operational objectives; (as specify a redundancy level, it is based on a security level (i.e., operational objectives that high level of security/safety) [0088] lines 8-11, achievement of operational objectives (e.g. high speed, high safety, low-latency, etc.); [0112] lines 11-20, some types of memory storage may provide varying levels of different operational characteristics that would be better suited for (a) certain types of data having certain types of data type characteristics; or (b) achieving a pre-determined operational objective as requested by, for example, the user or system administrator. These operational characteristics and operational objectives may include, but are not limited to, characteristics relating to speed, integrity, redundancy, persistence, security, methodology of implementing memory instructions).
It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to have combined the teaching of Schmidt, Dormitzer and Lacombe with Warfield because Warfield’s teaching of specifying the redundancy level based on meeting one or more operational objectives (i.e., safety/security) would have provided Schmidt, Dormitzer and Lacombe’s system with the advantage and capability to allow the system to keeping the data safe which improving the system reliability and performance (see Warfield, [0003] “keep data safe, and allow it to be accessed with excellent performance characteristics”).
Claims 11-12 are rejected under 35 U.S.C. 103 as being unpatentable over Schmidt and Dormitzer, as applied to claim 10 above, and further in view of SUBRAMANIAN et al. (US Pub. 2020/0104184 A1) and Lacombe et al. (US Patent. 6,055,647).
SUBRAMANIAN and Lacombe were cited in the previous Office Action.
As per claim 11, Schmidt and Dormitzer teach the invention according to claim 10 above. Schmidt further teaches to use machine learning to profile workloads, and determine a configuration for resources allocated to perform the second operations (Schmidt, [0034] lines 1-18, Information pertaining to the behavior of applications, as represented by system or library calls…pertaining to the application's behavior can be represented by a counting vector that counts the number of occurrences of each call during a pre-determined time period. Such pre-determined time period can be linked to time periods during which process resource usage metrics are measured; [0051] lines 22-26, Each application behavior monitor and each application scheduler is a set of processor executable instructions stored at a processor readable memory, e.g. one of memories 604A and 604N, and configured to be executed by a processor, e.g. one of CPUs 602A and 602N (as Infrastructure Processing Unit that utilizing both behavior monitor and each application scheduler); [0045] lines 1-6, the fixed-dimension vector of system load measurements and the fixed-dimension vector representation of the system event data set are provided as input to a machine learning model that has previously been trained to generate resource usage predictions for each of one or more applications during a future time period; [0040] lines 11-15, To provide a simple model for the use case of resource usage prediction, an LSTM can be trained with collected and preprocessed data that includes past usage statistics and past system call data. The learning of system call embeddings can be performed as a preprocessing step or within an end-to-end architecture; [0050] lines 6-11, At 530, the process uses the model for predicting future resource usage to predict resource usage for one or more application during a future time period. At 540, the process allocates resources (as to provide configuration of resources) within the computing system according to the resource usage predicted at 530 and so as to achieve a certain performance target, e.g. minimize average waiting time, provided by a value function having parameters Q. Allocating resources at 540 can be performed on a per-application basis as well as at other levels of granularity; please note: the resource modification was taught by Dormitzer).
Schmidt and Dormitzer fail to specifically teach wherein the network interface device is to use machine learning to profile workloads.
However, SUBRAMANIAN teaches wherein the network interface device is to use machine learning to profile workloads (SUBRAMANIAN, Fig. 5, 520 accelerator, 522 resource determination, suggested resource configuration to pod manager; [0044] lines 10-14, Resource determination module 522 can use an AI or ML model dedicated to determining and suggesting resource allocations for particular categories of applications (as to profile workloads (resource allocation sug