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 applicant’s communication filed December 19, 2024 in response to PTO Office Action dated September 28, 2024 The applicant’s remarks and amendment to the specification and/or claims were considered with the results that follow.
Claims 1-9 and 11-21 have been presented for examination in this application. In response to the last Office Action, claims 1, 11, and 15 have been amended. Claim 21 has been added.
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) 1-9 and 11-20 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Pan (US# 2012/0084497) in view of Chien et al. (Characterizing Deep-Learning I/O Workloads in TensorFlow).
Regarding claim 1, Pan teaches a method comprising:
… using training data comprising historical accesses … from during previous executions of the artificial intelligence application, of data objects from one or more … storage resources of a … storage platform … [0029-0030; event based prefetch list including non-sequential data locations, where the access patterns could be any grouping of storage locations in the list, or the entire list, as the list is based off a defined event that previously occurred [0030],
determining, …, subsequent patterns of access … of the data objects [event based prefetch list, 0030; see also para 0031, lines 1-5];
determining, based on the subsequent patterns of access …, a list of storage locations at the … storage platform for content expected to be used during a future execution of the … application; and prefetching, based on the list of storage locations, one or more data objects from a guided readahead from the list of storage locations at the cloud-based storage platform [guided non-sequential interpreted as non-sequential according to applicant’s Specification 000231, which comprises prefetching from multiple locations not necessarily at the same time] [Figure 2; 0033].
Pan teaches specific prefetching based on predefined events [0037] corresponding to a batch of data objects. However, Pan fails to clearly teach that the application is an artificial intelligence application and model, where the model has been trained on historical trends of data access for previous execution of an artificial intelligence application. Chien et al. teaches deep-learning (Artificial Intelligence) by way of the TensorFlow application’s prefetcher [page 2, line 17 – page 3, line 10]. The TensorFlow dataset has been trained using previous batches [page 1, section A, paragraph 2], such that the prefetching applications of deep learning/AI Tensorflow use upstream operation for prefetching [page 2, paragraph 2], which also applies to “..wherein the artificial intelligence application has previously accessed data objects in storage locations (prefetching) in the one or more cloud-based storage resources (Chien Tensorflow).
Pan also fails to teach that the storage platform is cloud-based. Chien teaches that the TensorFlow system functions with cloud-based storage systems [page 2, section II, second paragraph; e.g. Amazon Web Service, Google Cloud Storage] Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the type of application of Pan to include the artificial intelligence learning app and modelling of TensorFlow and associated cloud storage because then the effective cost of I/O on overall performance can be eliminated [page 1, lines 11-12].
Regarding claim 2, Pan teaches storing content that includes the one or more data objects within a memory accessible to the artificial intelligence application (Chien), wherein the content corresponds to the list of storage locations. [prefetched data stored in cache, 0033].
Regarding claim 3, Pan teaches further determining the patterns of access for a database, a system program (Chien, Abs.), or a user application (event) [Abstract; 0016, lines 1-8].
Regarding claim 4, Pan teaches the list of storage locations includes storage locations that are nonsequential. [0034, lines 1-5].
Regarding claim 5, Pan teaches wherein one or more addresses of the list of storage locations are a sequential increment from a previous address of the list of storage locations [0043; event prefetch list is sequentially followed to prefetch all associated event data].
Regarding claim 6, Chien et al. teaches wherein the patterns of access are based on historical trends of data access for previous executions of the artificial intelligence application or for previous executions of artificial intelligence applications that are similar to the artificial intelligence application. [Page 2, Prefetching].
Regarding claim 7, Pan teaches associating metadata with the list of storage locations, wherein the metadata includes one or more of: an application type, a user identification, a priority level [PrefetchCacheHitThreshold, indicating change in priority of prefetch list] an application name [EventIdentity], time of application use, or date of application use [Figure 4].
Regarding claim 8, Chien et al. teaches wherein determining the patterns of access is based on a second artificial intelligence application (TensorFlow applied in previous iteration) trained on one or more storage location accesses from a previous execution of the artificial intelligence application or of a similar artificial intelligence application [Page 2, line 17 – page 3, line 10].
Regarding claim 9, Chien et al. teaches wherein prefetching includes issuing an operating system call to a read ahead (prefetching)routine, and wherein the read ahead routine is a Linux system call [Page IV, section Tegner, line 3: CentOS operating system is operating system].
Regarding claim 11, Pan teaches a system comprising:
one or more storage systems comprising, respectively, one or more storage devices [Fig. 1; 190] and
one or more graphical processing units (GPUs), wherein the one or more GPUs are configured to communicate with the one or more systems [col. 67, lines 1-4; based on prefetch requests disclosed supra] over a communication fabric [while Pan does not explicitly recite GPUs, Pan teaches that a user will use a monitor or display device, which inherently requires the use of a processor able to process graphical data; 0067, lines 15-17], the one or more GPUs comprising a processing device [ GPUs are processing devices by definition] configured to
… using training data comprising historical accesses … from during previous executions of the artificial intelligence application, of data objects from one or more … storage resources of a … storage platform … [0029-0030; event based prefetch list including non-sequential data locations, where the access patterns could be any grouping of storage locations in the list, or the entire list, as the list is based off a defined event that previously occurred [0030],
determine, …, subsequent patterns of access … of the data objects [event based prefetch list, 0030; see also para 0031, lines 1-5];
determine, based on the subsequent patterns of access …, a list of storage locations at the … storage platform for content expected to be used during a future execution of the … application; and prefetch, based on the list of storage locations, one or more data objects from a guided readahead from the list of storage locations at the cloud-based storage platform [guided non-sequential interpreted as non-sequential according to applicant’s Specification 000231, which comprises prefetching from multiple locations not necessarily at the same time] [Figure 2; 0033].
Pan teaches specific prefetching based on predefined events [0037] corresponding to a batch of data objects. However, Pan fails to clearly teach that the application is an artificial intelligence application and model, where the model has been trained on historical trends of data access for previous execution of an artificial intelligence application. Chien et al. teaches deep-learning (Artificial Intelligence) by way of the TensorFlow application’s prefetcher [page 2, line 17 – page 3, line 10]. The TensorFlow dataset has been trained using previous batches [page 1, section A, paragraph 2], such that the prefetching applications of deep learning/AI Tensorflow use upstream operation for prefetching [page 2, paragraph 2], which also applies to “..wherein the artificial intelligence application has previously accessed data objects in storage locations (prefetching) in the one or more cloud-based storage resources (Chien Tensorflow).
Pan also fails to teach that the storage platform is cloud-based. Chien teaches that the TensorFlow system functions with cloud-based storage systems [page 2, section II, second paragraph; e.g. Amazon Web Service, Google Cloud Storage] Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the type of application of Pan to include the artificial intelligence learning app and modelling of TensorFlow and associated cloud storage because then the effective cost of I/O on overall performance can be eliminated [page 1, lines 11-12].
Regarding claim 12, Pan teaches storing content that includes the one or more data objects within a memory accessible to the artificial intelligence (Chien et al. ) application, wherein the content corresponds to the list of storage locations [prefetched data stored in cache, 0033].
Regarding claim 13, Pan teaches to determine the set of access patterns for a database, a system program (Chien, Abs.), or a user application (event) [Abstract; 0016, lines 1-8].
Regarding claim 14, Pan teaches wherein the list of storage locations includes storage locations that are nonsequential [0034, lines 1-5].
Claim 15 recites language very similar to that of claim 1 and 11, and is rejected for the same reasons as claim 1 and/or 11.
Claims 16-19 recite language very similar to that of claims 6-10, and are rejected for the same reasons as claims 6-9.
Regarding claim 20, Pan and Chien teach wherein the prefetching includes a hardware level controller accessing the list of storage locations [0030; pre-fetch controller (162) responsible to data retrieval].
Regarding claim 21, Pan and Chien teach further comprising training the artificial intelligence model including training the artificial intelligence model to perform one or more machine learning tasks received from a cloud-based artificial intelligence (AI) service [TensorFlow systems function with cloud-based storage systems [page 2, section II, second paragraph; e.g. Amazon Web Service, Google Cloud Storage] such that tasks {e.g., I/O] are performed in coordination with the TensorFlow AI system.
Response to Arguments
Applicant's arguments filed October 27, 2025 have been fully considered but they are not persuasive in view of the new grounds of rejection. The claims have been amended in such a way as to be similar to that of the March 17, 2025 filing.
Regarding Applicant’s argument of pages 9-10 that that Chien does not remedy the deficiencies of Pan, and is silent regarding the newly amended portions of the independent claims, the Examiner respectfully disagrees. Pan teaches determining patterns of access from previous events for the prefetching of data for future use. Chien has been included to teach the cloud based storage platform by way of the TensorFlow system which incorporates deep-learning (artificial intelligence), where each dataset is trained using previous batches (previously accessed data objects. The amended portions incorporate the idea of artificial intelligence (taught by Chien) into prefetching data from storage locations (cloud, also taught by Chien), which is well known to one of ordinary skill in the art. The “guided readahead” is not claimed to be specific to any single operation, and in the context of Applicant’s Specification 000231 the non-sequential readaheads (prefetches) may be done consecutively or separately. Both Pan and Chien teach prefetching data for the benefit of reducing I/O load and data retrieval. As disclosed above, Chien teaches using TensorFlow for prefetching from upstream operations for future use. Applicant’s claim language of “historical trends” is not defined in the claims and is therefore given a broad and reasonable interpretation of the phrase.
Regarding Applicant’s arguments that the “…prefetched one or more data objects is clearly distinct from "data objects from one or more cloud-based storage resources of a cloud-based storage platform" used "to generate a trained artificial intelligence model" and the same "data objects" using the trained artificial intelligence model to determine "subsequent patterns of access", the Examiner respectfully disagrees. The claim language does not clearly and positively recite that the two recitations of data objects are distinct and different, nor does the claims recite a timeline that prevents the set of data objects from being used at different time in different ways.
Therefore claims 1-9 and 10-21 remain rejected for the reasons disclosed supra.
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.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Brian R. Peugh whose telephone number is (571) 272-4199. The examiner can normally be reached on Monday-Friday from 7:30am to 3:30pm.
If attempts to reach the examiner by telephone are unsuccessful, the examiner's supervisor, Rocio Del Mar Perez-Velez, phone number 571-270-5935, can be reached. The fax phone number for the organization where this application or proceeding is assigned is 703-872-9306.
Any inquiry of a general nature or relating to the status of this application or proceeding should be directed to the receptionist whose telephone number is 571-272-2100.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
/BRIAN R PEUGH/Primary Examiner, Art Unit 2133