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 .
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) 1-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Goh(US 2014/0068627), Ben-Yehuda(US 2009/0083737) and Stark(US 2023/0118621).
Regarding claims 1, Goh discloses a method for reducing power consumption by a system including a digital unit, comprising: generating a workload signature of an incoming workload; matching the incoming workload, based on the workload signature, with a workload profile of stored workload profiles, wherein the digital unit executes compute operations for the incoming workload in a digital domain(Paragraphs 31, 32, 35; Data center 1 includes a plurality of compute nodes 2 and In response to the workload manager API 19 entering a new job into job queue 7, the policy engine 6 reviews the job type(e.g., streaming workload type) and the system management center 5 loads an optimal profile stored in the application profile analytics data base 8 into the policy engine 6).
Goh does not specifically disclose uniquely identifying the incoming workload. However, Ben-Yehuda discloses a workload signature of signatures 109 may be determined; sampling a plurality of values of the at least one parameter; determining one or more statistical values corresponding to the plurality of sampled values; and determining the workload signature(Paragraph 43) and system 100 and/or a management system (not shown) associated with system 100, may allocate resources to software service 102 based on the runtime workload classification, optionally taking into account any additional information in the annotations corresponding to the workload signature. System 100 and/or the management system may implement any suitable resource allocation and/or resource management mechanism, algorithm, and/or method to allocate the resources to software service 102(Paragraph 46). It would have been obvious to combine the teachings of Goh and Ben-Yehuda and Stark before the effective filing date of the claimed invention to a person having ordinary skill in the art to uniquely identifying the incoming workload. The motivation to do so would be to allows more accurate resource allocation based on more accurate workload identification.
Goh and Ben-Yehuda do not specifically disclose an optical unit, and selectively sending, a request to execute the incoming workload to either the optical unit or the digital unit based on the matching, wherein the optical unit executes compute operations for the incoming workload in an optical domain. However, Goh teaches data center 1 includes a plurality of compute nodes 2(Paragraph 31) and allocating resources(Paragraph 27). Stark discloses optical computing can potentially overcome this bottleneck as it inherently allows large bandwidth, massive parallelism, and low latency. Furthermore, it may drastically improve the energy efficiency of computations, thereby reducing the environmental impact of AI workloads. Integrated silicon photonics has majored as a technology over the last decades and now reaches excellent technological readiness. It provides the means for economical large-scale fabrication and can be integrated with electronic systems with reasonable overhead, hence enabling a general adaption of optical computing in data centers(Paragraph 19). It would have been obvious to combine the teachings of Goh, Ben-Yehuda and Stark before the effective filing date of the claimed invention to a person having ordinary skill in the art to include an optical unit, and selectively sending, a request to execute the incoming workload to either the optical unit or the digital unit based on the matching, wherein the optical unit executes compute operations for the incoming workload in an optical domain. The motivation to do so would be to allows large bandwidth, massive parallelism, and low latency. Furthermore, it may drastically improve the energy efficiency of computations, thereby reducing the environmental impact of AI workloads(Stark: Paragraph 19).
Regarding claim 2, Goh, Ben-Yehuda and Stark disclose the method of claim 1, wherein the profile comprises: the workload signature; a first set of performance measures associated with executing the workload in the digital unit; and a second set of performance measures associated with executing the workload in the optical unit(Goh: Paragraphs 35-36, Identifying a workload type. Policy rules corresponding to power thresholds and workload profiles are established at step 301. The policy rules are set corresponding to platform type and workload profile. As mentioned above, this may be based on historical data, configuration data set by an administrator, by a combination thereof, or be based on other metrics. It would have been obvious to include performance measurements of the digital unit and the optical unit so that the optimal performance can be achieved).
Regarding claim 3, Goh, Ben-Yehuda and Stark disclose the method of claim 2, wherein at least one of the first set of performance measures, or the second set of performance measures, is a level of power consumed when executing respective workload(Goh: Paragraph 35, Power consumption and performance of the workload type on the platform type are measured in step 205).
Regarding claim 4, Goh, Ben-Yehuda and Stark disclose the method of claim 2, wherein the sending, comprises: scheduling the incoming workload to be executed in the optical unit responsive to a second set of performance measures of the associated profile indicated as outperforming the first set of performance measures of the associated profile(Goh: Paragraphs 32, 35; In response to the workload manager API 19 entering a new job into job queue 7, the policy engine 6 reviews the job type(e.g., streaming workload type) and the system management center 5 loads an optimal profile stored in the application profile analytics data base 8 into the policy engine 6).
Regarding claim 5, Goh, Ben-Yehuda and Stark disclose the method of claim 2, further comprising: generating the workload signature of the profile by: monitoring, by a trace capture unit of the system, transactions, including a task submission transaction associated with the workload; extracting information out of the transactions, the extracted information is characteristic of the workload; and generating, based on the extracted information, the workload signature(Goh: Paragraph 35, measuring the power consumption and performance of a workload running on a platform type in step 201. Next, characteristics of the workload are identified in step 202, and a workload type is assigned to the workload in step 203. For example, if the workload includes large serialized streams of data, the workload type could be a streaming workload type.)
Regarding claim 6, Goh, Ben-Yehuda and Stark disclose the method of claim 2, further comprising: generating the first set of performance measures of the profile by: monitoring, by the trace capture unit, a first group of transactions associated with the workload, including a task submission transaction and a respective task completion transaction that are directed at the digital unit of the system; and computing, based on information extracted from the first group of transactions, the first set of performance measures(Goh: Paragraph 35, measuring the power consumption and performance of a workload running on a platform type in step 201. Next, characteristics of the workload are identified in step 202, and a workload type is assigned to the workload in step 203. For example, if the workload includes large serialized streams of data, the workload type could be a streaming workload type.)
Regarding claim 7, Goh, Ben-Yehuda and Stark disclose the method of claim 2, further comprising: generating the second set of performance measures of the profile by: monitoring, by the trace capture unit, a second group of transactions associated with the workload, including a task submission transaction and a respective task completion transaction that are directed at the optical unit; and computing, based on information extracted from the second group of transactions, the second set of performance measures(Goh: Paragraph 35, measuring the power consumption and performance of a workload running on a platform type in step 201. Next, characteristics of the workload are identified in step 202, and a workload type is assigned to the workload in step 203. For example, if the workload includes large serialized streams of data, the workload type could be a streaming workload type.)
Regarding claim 8, Goh, Ben-Yehuda and Stark disclose the method of claim 1, further comprising: encoding a data word sequence fed to a converter of the optical unit, wherein the encoding reduces the number of transitions across corresponding bits in successive data words of the data word sequence; and converting the encoded data word sequence by the converter, wherein when the converter is a digital-to-analog converter the encoding is by a digital encoder and when the converter is an analog-to-digital converter the encoding is by an analog encoder(Stark: Paragraph 42, The unit 10, transmitter 11, receiver 13, and data reconstruction unit 14 involved in the present apparatus 1 are independently known, as such, in the prior art. They merely need to be connected and configured to serve the present purpose, i.e., provide initial data, convert them to a 1D data stream encoded as an optical signal(analog to digital converter). Further, minimizing bit transitions is well known concept and would be obvious to one of ordinary skill in the art to encode the words to minimize bit transitions for the motivation of reducing power).
Regarding claim 9, Goh, Ben-Yehuda and Stark disclose the method of claim 1, further comprising: generating signatures of respective incoming workloads; associating the incoming workloads, based on the signatures, with respective profiles of workload profiles; buffering task submission transactions, each of the transactions represents a request to execute a respective workload of the incoming workloads; merging the buffered task submission transactions into a merged task submission transaction; and sending the merged task submission transaction to the optical unit(Goh: Paragraphs 32, 35; In response to the workload manager API 19 entering a new job into job queue 7, the policy engine 6 reviews the job type(e.g., streaming workload type) and the system management center 5 loads an optimal profile stored in the application profile analytics data base 8 into the policy engine 6).
Regarding claim 10, Goh, Ben-Yehuda and Stark disclose the method of claim 1, wherein: before the sending, transitioning a converter of the optical unit from a sleep mode into an operational mode(Goh: Paragraph 24, one or more policy rules that correspond to one or more parametrics in the data center are setup by utilizing the historical tracking and analysis data analysis data for software applications running on hardware resources in the data center, or a combination thereof. Parametrics monitored by the method include, yet are not limited to, those that relate to resiliency, power consumption, power balancing, power management, maintenance, error rate, and performance. It would have been obvious to wake up components that are going to be used).
Regarding claim 11, Goh discloses a system, including a digital unit, for reducing power consumption, comprising: at least one processor; and memory storing instructions that, when executed by the at least one processor(Paragraph 11, Some embodiments of the data center have a plurality of processors and shared memory storing one or more programs for execution by the plurality of processors ), cause the system to: generate a workload signature of an incoming workload, match the incoming workload, based on the workload signature, with a workload profile of stored workload profiles, wherein the digital unit executes compute operations for the incoming workload in a digital domain(Paragraphs 31, 32, 35; Data center 1 includes a plurality of compute nodes 2 and In response to the workload manager API 19 entering a new job into job queue 7, the policy engine 6 reviews the job type(e.g., streaming workload type) and the system management center 5 loads an optimal profile stored in the application profile analytics data base 8 into the policy engine 6).
Goh does not specifically disclose uniquely identifying the incoming workload. However, Ben-Yehuda discloses a workload signature of signatures 109 may be determined; sampling a plurality of values of the at least one parameter; determining one or more statistical values corresponding to the plurality of sampled values; and determining the workload signature(Paragraph 43) and system 100 and/or a management system (not shown) associated with system 100, may allocate resources to software service 102 based on the runtime workload classification, optionally taking into account any additional information in the annotations corresponding to the workload signature. System 100 and/or the management system may implement any suitable resource allocation and/or resource management mechanism, algorithm, and/or method to allocate the resources to software service 102(Paragraph 46). It would have been obvious to combine the teachings of Goh and Ben-Yehuda and Stark before the effective filing date of the claimed invention to a person having ordinary skill in the art to uniquely identifying the incoming workload. The motivation to do so would be to allows more accurate resource allocation based on more accurate workload identification.
Goh and Ben-Yehuda do not specifically disclose an optical unit, and selectively sending, a request to execute the incoming workload to either the optical unit or the digital unit based on the matching, wherein the optical unit executes compute operations for the incoming workload in an optical domain. However, Goh teaches data center 1 includes a plurality of compute nodes 2(Paragraph 31) and allocating resources(Paragraph 27). Stark discloses optical computing can potentially overcome this bottleneck as it inherently allows large bandwidth, massive parallelism, and low latency. Furthermore, it may drastically improve the energy efficiency of computations, thereby reducing the environmental impact of AI workloads. Integrated silicon photonics has majored as a technology over the last decades and now reaches excellent technological readiness. It provides the means for economical large-scale fabrication and can be integrated with electronic systems with reasonable overhead, hence enabling a general adaption of optical computing in data centers(Paragraph 19). It would have been obvious to combine the teachings of Goh, Ben-Yehuda and Stark before the effective filing date of the claimed invention to a person having ordinary skill in the art to include an optical unit, and selectively sending, a request to execute the incoming workload to either the optical unit or the digital unit based on the matching, wherein the optical unit executes compute operations for the incoming workload in an optical domain. The motivation to do so would be to allows large bandwidth, massive parallelism, and low latency. Furthermore, it may drastically improve the energy efficiency of computations, thereby reducing the environmental impact of AI workloads(Stark: Paragraph 19).
Regarding claim 12, Goh, Ben-Yehuda and Stark disclose the system of claim 11, wherein the profile comprises: the workload signature; a first set of performance measures associated with executing the workload in the digital unit; and a second set of performance measures associated with executing the workload in the optical unit(Goh: Paragraphs 35-36, Identifying a workload type. Policy rules corresponding to power thresholds and workload profiles are established at step 301. The policy rules are set corresponding to platform type and workload profile. As mentioned above, this may be based on historical data, configuration data set by an administrator, by a combination thereof, or be based on other metrics. It would have been obvious to include performance measurements of the digital unit and the optical unit so that the optimal performance can be achieved).
Regarding claim 13, Goh, Ben-Yehuda and Stark disclose the system of claim 12, wherein a performance measure, of the first set or the second set, is a level of power consumed by the execution of the workload or an execution time of the workload(Goh: Paragraph 35, Power consumption and performance of the workload type on the platform type are measured in step 205).
Regarding claim 14, Goh, Ben-Yehuda and Stark disclose the system of claim 12, wherein the sending, comprises: scheduling the incoming workload to be executed in the optical unit responsive to a second set of performance measures of the associated profile indicated as outperforming the first set of performance measures of the associated profile(Goh: Paragraphs 32, 35; In response to the workload manager API 19 entering a new job into job queue 7, the policy engine 6 reviews the job type(e.g., streaming workload type) and the system management center 5 loads an optimal profile stored in the application profile analytics data base 8 into the policy engine 6).
Regarding claim 15, Goh, Ben-Yehuda and Stark disclose the system of claim 12, wherein the instructions further cause the system to: generate the workload signature of the profile by: monitoring, by a trace capture unit of the system, transactions, including a task submission transaction associated with the workload; extracting information out of the transactions, the extracted information is characteristic of the workload; and generating, based on the extracted information, the workload signature(Goh: Paragraph 35, measuring the power consumption and performance of a workload running on a platform type in step 201. Next, characteristics of the workload are identified in step 202, and a workload type is assigned to the workload in step 203. For example, if the workload includes large serialized streams of data, the workload type could be a streaming workload type.)
Regarding claim 16, Goh, Ben-Yehuda and Stark disclose the system of claim 12, wherein the instructions further cause the system to: generate the first set of performance measures of the profile by: monitoring, by the trace capture unit, a first group of transactions associated with the workload, including a task submission transaction and a respective task completion transaction that are directed at the digital unit of the system; and computing, based on information extracted from the first group of transactions, the first set of performance measures(Goh: Paragraph 35, measuring the power consumption and performance of a workload running on a platform type in step 201. Next, characteristics of the workload are identified in step 202, and a workload type is assigned to the workload in step 203. For example, if the workload includes large serialized streams of data, the workload type could be a streaming workload type.)
Regarding claim 17, Goh, Ben-Yehuda and Stark disclose the system of claim 12, wherein the instructions further cause the system to: generate the second set of performance measures of the profile by: monitoring, by the trace capture unit, a second group of transactions associated with the workload, including a task submission transaction and a respective task completion transaction that are directed at the optical unit; and computing, based on information extracted from the second group of transactions, the second set of performance measures(Goh: Paragraph 35, measuring the power consumption and performance of a workload running on a platform type in step 201. Next, characteristics of the workload are identified in step 202, and a workload type is assigned to the workload in step 203. For example, if the workload includes large serialized streams of data, the workload type could be a streaming workload type.)
Regarding claim 18, Goh, Ben-Yehuda and Stark disclose the system of claim 11, wherein the instructions further cause the system to: encode a data word sequence fed to a converter of the optical unit, wherein the encoding reduces the number of transitions across corresponding bits in successive data words of the data word sequence; and converting the encoded data word sequence by the converter, wherein when the converter is a digital-to-analog converter the encoding is by a digital encoder and when the converter is an analog-to-digital converter the encoding is by an analog encoder(Stark: Paragraph 42, The unit 10, transmitter 11, receiver 13, and data reconstruction unit 14 involved in the present apparatus 1 are independently known, as such, in the prior art. They merely need to be connected and configured to serve the present purpose, i.e., provide initial data, convert them to a 1D data stream encoded as an optical signal(analog to digital converter). Further, minimizing bit transitions is well known concept and would be obvious to one of ordinary skill in the art to encode the words to minimize bit transitions for the motivation of reducing power).
Regarding claim 19, Goh, Ben-Yehuda and Stark disclose the system of claim 11, wherein the instructions further cause the system to: generate signatures of respective incoming workloads; associate the incoming workloads, based on the signatures, with respective profiles of workload profiles; buffer task submission transactions, each of the transactions represents a request to execute a respective workload of the incoming workloads; merge the buffered task submission transactions into a merged task submission transaction; and send the merged task submission transaction to the optical unit, wherein before the sending, a converter of the optical unit is transitioned from a sleep mode into an operational mode(Goh: Paragraph 24, one or more policy rules that correspond to one or more parametrics in the data center are setup by utilizing the historical tracking and analysis data analysis data for software applications running on hardware resources in the data center, or a combination thereof. Parametrics monitored by the method include, yet are not limited to, those that relate to resiliency, power consumption, power balancing, power management, maintenance, error rate, and performance. It would have been obvious to wake up components that are going to be used).
Regarding claim 20, Goh discloses non-transitory computer-readable medium comprising instructions executable by at least one processor to perform a method for reducing power consumption by a system including a digital unit and an optical unit, the method comprising: generating a workload signature of an incoming workload; matching the incoming workload, based on the workload signature, with a workload profile of stored workload profiles(Paragraph 11, Some embodiments of the data center have a plurality of processors and shared memory storing one or more programs for execution by the plurality of processors), cause the system to: generate a workload signature of an incoming workload, match the incoming workload, based on the workload signature, with a workload profile of stored workload profiles, wherein the digital unit executes compute operations for the incoming workload in a digital domain(Paragraphs 31, 32, 35; Data center 1 includes a plurality of compute nodes 2 and In response to the workload manager API 19 entering a new job into job queue 7, the policy engine 6 reviews the job type(e.g., streaming workload type) and the system management center 5 loads an optimal profile stored in the application profile analytics data base 8 into the policy engine 6).
Goh does not specifically disclose uniquely identifying the incoming workload. However, Ben-Yehuda discloses a workload signature of signatures 109 may be determined; sampling a plurality of values of the at least one parameter; determining one or more statistical values corresponding to the plurality of sampled values; and determining the workload signature(Paragraph 43) and system 100 and/or a management system (not shown) associated with system 100, may allocate resources to software service 102 based on the runtime workload classification, optionally taking into account any additional information in the annotations corresponding to the workload signature. System 100 and/or the management system may implement any suitable resource allocation and/or resource management mechanism, algorithm, and/or method to allocate the resources to software service 102(Paragraph 46). It would have been obvious to combine the teachings of Goh and Ben-Yehuda and Stark before the effective filing date of the claimed invention to a person having ordinary skill in the art to uniquely identifying the incoming workload. The motivation to do so would be to allows more accurate resource allocation based on more accurate workload identification.
Goh and Ben-Yehuda do not specifically disclose an optical unit, and selectively sending, a request to execute the incoming workload to either the optical unit or the digital unit based on the matching, wherein the optical unit executes compute operations for the incoming workload in an optical domain. However, Goh teaches data center 1 includes a plurality of compute nodes 2(Paragraph 31) and allocating resources(Paragraph 27). Stark discloses optical computing can potentially overcome this bottleneck as it inherently allows large bandwidth, massive parallelism, and low latency. Furthermore, it may drastically improve the energy efficiency of computations, thereby reducing the environmental impact of AI workloads. Integrated silicon photonics has majored as a technology over the last decades and now reaches excellent technological readiness. It provides the means for economical large-scale fabrication and can be integrated with electronic systems with reasonable overhead, hence enabling a general adaption of optical computing in data centers(Paragraph 19). It would have been obvious to combine the teachings of Goh, Ben-Yehuda and Stark before the effective filing date of the claimed invention to a person having ordinary skill in the art to include an optical unit, and selectively sending, a request to execute the incoming workload to either the optical unit or the digital unit based on the matching, wherein the optical unit executes compute operations for the incoming workload in an optical domain. The motivation to do so would be to allows large bandwidth, massive parallelism, and low latency. Furthermore, it may drastically improve the energy efficiency of computations, thereby reducing the environmental impact of AI workloads(Stark: Paragraph 19).
Response to Arguments
Applicant’s arguments with respect to claim 1-20 have been considered, but are moot due to new grounds of rejection.
Conclusion
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.
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 NIMESH G PATEL whose telephone number is (571)272-3640. The examiner can normally be reached Monday-Friday, 8:15-4:15.
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, Jaweed Abbaszadeh can be reached on 571-270-1640. 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.
/NIMESH G PATEL/Primary Examiner, Art Unit 2187
Prosecution Insights