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 application filed on 11/8/2024, wherein claims 1-15 are pending, claiming priority based on foreign priority of CN202410121566.3, dated 1/29/2024.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 1-15 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
The following claim terms are unclear and indefinite:
As for claim 1, it is unclear what is meant by “designated virtual card” because the Specification appears to describe the claimed term with 2 different means that lead to entirely different operations. Under a first construction, “designated” is a label for a card that has been found to have no time slice allocated and have a non-empty load waiting for execution (specification, paragraph 35, “sequentially traverse the virtual cards that have no corresponding time slice allocated…if there is a virtual card having a non-empty load, determine the virtual card as a designated virtual card…” and paragraph 31, “…if there has been always no virtual card having a non-empty load and no time slice allocated in the virtual card set, then allocating no corresponding time slice…”) In this instance, the term designated imposes no limitations, and the conditions are merely criteria by which the card was labeled. Under a second construction, the term “designated” has its ordinary meaning of a card selected or appointed already/ahead of time/in advance. (Specification, paragraph 31, “detecting in real time or periodically whether there is a designated virtual card having a non-empty load and no time slice allocated in the virtual card set...if not, early ending the scheduling period.” To sequentially traverse the virtual cards in the virtual card set that has no corresponding time slice allocated, to determine whether there is a designated virtual card having a non-empty load…”). In this second construction, the period can end while an undesignated card that has a non-empty load and no allocated slice can remain pending, and the scope of the claim depends on the basis of which the designation is made, which the claim does not recite, and the specification does not describe. This in contrast to the current application’s specification’s description of high-priority virtual card, which states “….whether there is a high-priority virtual card (i.e., a preset virtual card of high-priority) in the virtual card set…” (paragraph 26, which explicitly states it’s a pre-configured status. Consequently, because the 2 different possible meanings of designated that can lead to two different sequences of operations, the claimed term is indefinite. For the purpose of examination, Examiner assume designated to mean any virtual card that has a workload and is not yet allocated time slice in the set of virtual cards claimed, not a pre-configured virtual card that is configured as a ”designated virtual card”.
As for claims 6 and 11, they contain similar defect as claim 1 above. Thus, they are rejected under the same rationales.
As for claims 2-5, 7-10, 12-15 they are rejected for failure to cure defects of the claim upon which they depend.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1, 4-6, 11, and 14-15 are rejected under 35 U.S.C. 103 as being unpatentable over Liu et al. (US PGPUB 2018/0349204), in view of Cheng (US PGPUB 2019/0004839).
As for claim 1, Liu teaches a method for scheduling a time slice (Abstract, “…implementing a virtual GPU including….allocating…virtual GPUs a running time slice…”), comprising:
determining a scheduling period based on a number of virtual cards in a virtual card set [virtual GPUs of Type C and D] and a maximum scheduling delay [maximum scheduling period] (paragraph 37, “…maximum scheduling period is a maximum allowable polling time of all virtual GPUs configured on the physical GPU. When multiple virtual GPUs are configured on one physical GPU, a sum of time...all the virtual GPUs can be referred to as a total scheduling period…..cannot exceed the maximum scheduling period…when the maximum scheduling period is 40 milliseconds (ms), four running time slices of 40ms, 20ms, 10ms, and 5 ms can be respectively used for the virtual GPUs of the four specifications A, B, C and D….” and paragraph 47, “…two virtual GPUs of the type C and one virtual GPU of the type D…time slice of the virtual GPU of the type C is 10ms, and the time slice of the virtual GPU of the type D is 5 ms. One scheduling period is 10+10+5=25ms. When scheduling lasts for 25ms, each virtual GPU has a running time corresponding to the time slice of the virtual GPU….one scheduling period of one physical GPU maybe less than the maximum scheduling period according to the quantity of virtual GPUs running on the physical GPU…” in view of paragraph 74-75, “…virtual GPU C2 can be deleted…the scheduling period for running virtual GPUs can be reduced and changed to 10+5=15ms…” teaching scheduling period is based on the number of vGPUs and the maximum scheduling delay, and when the number of vGPUs change, the period is redetermined.);
determining, based on the scheduling period, time slices corresponding to the virtual cards [virtual GPUs] in the virtual card set, a time slice being used to indicate a duration of using an external computing power resource by a virtual card (paragraph 31, “…each of the virtual GPUs can be allocated with a running time slice corresponding to the resource requirement of the virtual GPU…”, paragraph 37, “…different types of virtual GPUs can be determined according to the percentage of the maximum scheduling period occupied by the virtual GPUs….when maximum …is 40ms, four running time slices of 40ms, 20ms, 10ms, and 5 ms can be respectively used for the virtual GPUs of the four specifications A, B, C, and D…” paragraph 73, “…sum of all the time slices …is less than the maximum scheduling period…virtual GPU C1 can occupy 40% of time slices of the physical GPU, virtual GPU C2 can occupy another 40% of the time slices of the physical GPU, and virtual GPU D can occupy 20% of the time slices of the physical GPU during the actual running of the virtual GPUs…” Thus, when the period changes the slices are re-derived from it “the 10ms released by virtual GPU C2 can be proportionally allocated to virtual GPUs C1 and D” (Paragraph 74-75)); and
performing a scheduling operation on the virtual card set (paragraph 41, “…a next read-to-run virtual GPU can be selected from a scheduling queue…” and paragraph 45, “…the virtual GPU can be stopped from running….when an accumulated running time of the virtual GPU reaches a running time slice allocated to the virtual GPU…”), and ending the scheduling period (paragraph 47, “…. selecting a next ready-to-run virtual GPU from a scheduling queue….when scheduling lasts for 25ms, each virtual GPU has a running time corresponding to the time slice of the virtual GPU, and the process enters a next scheduling period after the running time elapsed…” teaches ending the scheduling period when no vGPU remains that has not yet had its slice.)
Examiner note “ending the scheduling period in response to While Liu teaches scheduling operation on the virtual card set and scheduling of virtual card having a non-empty load (i.e., ready to run), and Liu additionally is capable of scheduling unassigned vGPUs with workload waiting for execution to the current scheduling period when there is sufficient resource available to execute such virtual cards (see, e.g., paragraph 52-53). Thus, it would be obvious to a person of ordinary skill in the art before the effective filing date of the application to recognize that when the current vGPUs finish its workload and there are no ready to run vGPUs waiting to be assigned a time slice, the scheduling period is ended to avoid wasting idle running of GPU. However, in the interest of compact prosecution, Examiner note Liu does not explicitly teach checking whether each not-yet-served vGPU has pending work and has not been allocated a time slice.
However, Cheng teaches a known method of time-sliced workload resource allocation on virtualized entities running on accelerated processing devices including ending the scheduling period in response to determining that there is no designated virtual entity having a non-empty load and no corresponding time slice allocated in the virtual card set (paragraphs 52, 61, “…performing the virtualization context switch early is to allow other functions to perform work when the current function has no more work to perform, instead of idling the APD 116 for the remainder of the time-slice for the current function, by performing a virtualization context switch to a subsequent function…skipping the next turn of the subsequent if that function…has not received any work since the early virtualization context switch was performed…” and “…after performing a virtualization context switch for a function, if it becomes the turn for a function again but no doorbells have been received, then …skips that function…” in view of paragraph 35, teaching determining if there is a remaining virtualized entity, that have no corresponding time slice allocated, has a load to process or not, and in response to determining a remaining virtualized entity that do not have a load to process, skip/do not schedule the virtualized entity. Examiner note the Specification teaching periodic time-division of physical card resource, wherein ending the scheduling period is merely recited with no mentioning of what happens after an early-ended period (paragraph 31, 44). Because the application is directed to a method is a periodic time-division allocation (paragraph 3), with no teaching of not only what happens following an ended scheduling period, and an additional branch stating if there is always no virtual card having a non-empty load and no time slice allocated in the virtual card set, then allocating no corresponding time slice to avoid wasting resources (paragraph 31). In another word, system can’t find work, don’t allocate a next time slice. Thus, BRI of “ending a scheduling time period” is understood to at least include not scheduling any additional work within the current time slice until the end of the time slice in response to not able to find a different workload that has work to do now. Here, not scheduling/skipping the next turn of the subsequent function when the next subsequent function haven’t received any work (i.e., empty-load), skip/do not scheduled in the current time slice functionally is ending the scheduling period in response to determining no virtual entity with non-empty load yet to be scheduled for a time slice because it similarly determined no entity with non-empty load yet to be scheduled and doesn’t run it). This known technique is applicable to the system of Liu as they both share characteristics and capabilities, namely, they are directed to time-sliced workload resource allocation on virtualized entities.
One of ordinary skill in the art before the effective filing date of the application would have recognized that applying the known technique of Cheng would have yielded predictable results and resulted in an improved system. It would have been recognized that applying the technique of Cheng to the teachings of Liu would have yielded predictable results because the level of ordinary skill in the art demonstrated by the references applied shows the ability to incorporate such data processing features into similar systems. Further, applying ending the scheduling period in response to determining that there is no designated virtual entity having a non-empty load and no corresponding time slice allocated in the virtual card set to Liu with ending the scheduling period accordingly, would have been recognized by those of ordinary skill in the art as resulting in an improved system that would improve resource utilization efficiency. (Cheng, paragraph 9).
As for claims 6 and 11, they contain similar limitations as claim 1 above. Thus, they are rejected under the same rationales.
As for claim 4, Liu also teaches in response to determining that in the virtual card set there is a designated virtual card having a non-empty load and no corresponding time slice allocated, determining a target virtual card in the designated virtual card and allocating a corresponding time slice to the target virtual card (paragraph 41-42. The next workload in the queue in the prior art is understood as having a non-empty load and not yet allocated time slice.); and in response to determining that a load of the target virtual card is empty, continuing to perform the scheduling operation before the time slice corresponding to the target virtual card runs out (paragraph 44-45 in view of paragraph 47, “…one scheduling period is 10+10+5=25ms. When the scheduling lasts for 25 ms, each virtual GPU has a running time corresponding to the time slice of the virtual GPU, and the process enters a next scheduling period after the running time elapses. Examiner note when the target virtual card (which is a designated virtual card) is allocated a time slice, the designated virtual card has a non-empty load, meaning, it has work, and the load is not empty. Thus, “…before the time slice corresponding to the target virtual card runs out, continuing to perform the above scheduling operation in response to determining that the load of the target virtual card is empty” can be understood as letting the time slice corresponding to the target virtual card continue to continue where the “above scheduling operation” refers to the scheduling operation of allocating a corresponding time slice to the target virtual card.).
In addition, Cheng also teaches determining that in the virtual entities there is a designated virtual entity having a non-empty load and no corresponding time slice allocated, determining a target virtual entity in the designated virtual entity and allocating a corresponding time slice to the target virtual entity (paragraphs 52 and 61 in view of paragraph 34 teaching selecting a targeted workload, where it skips ones with no actual load ready to execute (i.e., it selects the ones with load to execute); and in response to determining that a load of the target virtual entity is empty, continuing to perform the scheduling operation before the time slice corresponding to the target virtual entity runs out (52 and 61. The system constantly monitors for idleness. Thus, prior art teaches “…before the time slice corresponding to the target virtual card runs out, continuing to performing the above scheduling operation in response to determining that the load of the target virtual card is empty” refers to a second interpretation of the claim language where it refers to the overall scheduling operation of repeated scheduling of next workload when the current workload one runs out of work (instead of the scheduling operation of claim 4 that schedules the target virtual card time slice, and thus can be interpreted as letting the said target virtual card time slice finish.)). Rationale to combine same as claim 1 above.
As for claims 14, they contain similar limitations as claim 4 above. Thus, they are rejected under the same rationales.
As for claim 5, Liu also teaches in response to determining that in the virtual card set there is a designated virtual card having a non-empty load and no corresponding time slice allocated, determining a target virtual card in the designated virtual card and allocating a corresponding time slice to the target virtual card (paragraphs 41-42); and continuing to perform the scheduling operation, in response to determining that the time slice corresponding to the target virtual card runs out (paragraph 44-45 in view of paragraph 47 thereby teaching letting the current time slice run out then repeat scheduling next workload)
In addition, Cheng teaches the same (paragraphs 52 and 61 in view of 34 teaching scheduling the target virtual card time slice early when current workload became idle (ran out of work), and paragraph 40 teaching letting time-slice for virtual entity run to expiration when it has work and non-idle). Rationale to combine same as claim 1 above.
As for claims 15, they contain similar limitations as claim 5 above. Thus, they are rejected under the same rationales.
Claim(s) 2-3, 7-10, and 12-13 are rejected under 35 U.S.C. 103 as being unpatentable over Liu et al. (US PGPUB 2018/0349204), in view of Cheng (US PGPUB 2019/0004839), in view of Chen et al. (US PGPUB 2003/0120705).
As for claim 2, Liu teaches in response to determining that there is no higher-priority virtual card in the card set the time slices corresponding to the virtual cards in the virtual card set based on the scheduling period and weight of the virtual cards (paragraph 36 and 47. Examiner note “in response to determining that there is no higher priority virtual card in the card set…” does not have an alternative determining of existence of higher priority virtual card in the card set, thus, the BRI of the claim includes when the subsequent time slice limitation is executed both when there is, and there is not a higher priority virtual card, and rendering it to have no effect on the remaining claim limitations as the remaining limitations can be understood as always executed.).
As noted above, Liu teaches the BRI of the claim because the time slice correspondence limitation can be executed regardless of the determining step. However, in the interest of compact prosecution, Examiner note that Liu and Cheng do not Liu and Cheng do not explicitly teach time slices corresponding to the virtual cards…based on scheduling period and weights of the virtual card is only performed when there is no high-priority virtual cards, and there is an alternative step when there is high-priority virtual cards.
However, Chen teaches a known method of resource allocation for shared network resources including in response to determining that there is no higher priority workloads in the workload set, the time slice is based on the scheduling period and weight of the workloads (paragraphs 23 and 26 and Fig. 1. “…weighted round robin scheme…” and “…all the classes are served in accordance with a smooth deficit weighted….less the priority assignment for one class…” Examiner note the weighted round robin is done except for the priority class.). This known technique is applicable to the system of Liu and Cheng as they both share characteristics and capabilities, namely, they are directed to shared resource allocation in a networked environment.
One of ordinary skill in the art before the effective filing date of the application would have recognized that applying the known technique of Chen would have yielded predictable results and resulted in an improved system. It would have been recognized that applying the technique of Chen to the teachings of Liu and Cheng would have yielded predictable results because the level of ordinary skill in the art demonstrated by the references applied shows the ability to incorporate such data processing features into similar systems. Further, applying in response to determining that there is no higher priority workloads in the workload set, the time slice is based on the scheduling period and weight of the workloads to Liu and Cheng with time slice is based on the scheduling period and weight of the virtual cards accordingly, would have been recognized by those of ordinary skill in the art as resulting in an improved system that would allow flexible implementation of different scheduling schemes within a single algorithm. (Chen, paragraph 12).
As for claims 7 and 12, they contain similar limitations as claim 2 above. Thus, they are rejected under the same rationales.
As for claim 3, Liu teaches resource allocation to virtual cards where resource are quantified based on size of time slices (paragraphs 37 and 47) and in response to determining that a load of the virtual card is empty and that there is a remaining time slice in the scheduling period, determining, based on the remaining time slice of the virtual cards in a sub-virtual card set, a time slice corresponding to a virtual card in the sub-virtual card set (paragraph 71).
Chen also teaches allocating, in response to determining that there is a high-priority workload in the workload set, the scheduling period to the high-priority workload (paragraph 22 teaching resource allocation to priority class one workloads until its empty, then releasing resources to the remaining workloads); and in response to determining that a load of the high-priority workload is empty and that there is a remaining time slice in the scheduling period, determining, based on the remaining time slice and weights of workloads in a sub-workload set, a time slice corresponding to a workload in the sub-workload set, wherein the sub-workload set is composed of non-high-priority workloads in the workload set (paragraph 22 teaching resource allocation to priority class one workloads until no load to process before releasing resources to the remaining workloads. Thus, when the lower priority workloads start to get assigned resources, the higher priority workload’s load is determined to be empty already. Examiner note, Liu already supplied the teaching of proportional reallocation of a remaining time slice of a virtual card/workload to sub-virtual card sets/workloads not including the virtual card whose remaining time slice is getting redistributed (Liu, paragraph 71).). Rationale to combine same as claim 2 above.
As for claims 8 and 13, they contain similar claim limitations as claim 3 above. Thus, they are rejected under the same rationales.
As for claim 9, Liu also teaches in response to determining that in the virtual card set there is a designated virtual card having a non-empty load and no corresponding time slice allocated, determining a target virtual card in the designated virtual card and allocating a corresponding time slice to the target virtual card (paragraph 41-42. The next workload in the queue in the prior art is understood as having a non-empty load and not yet allocated time slice.); and in response to determining that a load of the target virtual card is empty, continuing to perform the scheduling operation before the time slice corresponding to the target virtual card runs out (paragraph 44-45 in view of paragraph 47, “…one scheduling period is 10+10+5=25ms. When the scheduling lasts for 25 ms, each virtual GPU has a running time corresponding to the time slice of the virtual GPU, and the process enters a next scheduling period after the running time elapses. Examiner note when the target virtual card (which is a designated virtual card) is allocated a time slice, the designated virtual card has a non-empty load, meaning, it has work, and the load is not empty. Thus, “…before the time slice corresponding to the target virtual card runs out, continuing to perform the above scheduling operation in response to determing that the load of the target virtual card is empty” can be understood as letting the time slice corresponding to the target virtual card continue to continue where the “above scheduling operation” refers to the scheduling operation of allocating a corresponding time slice to the target virtual card.).
In addition, Cheng also teaches determining that in the virtual entities there is a designated virtual entity having a non-empty load and no corresponding time slice allocated, determining a target virtual entity in the designated virtual entity and allocating a corresponding time slice to the target virtual entity (paragraphs 52 and 61 in view of paragraph 34 teaching selecting a targeted workload, where it skips ones with no actual load ready to execute (i.e., it selects the ones with load to execute); and in response to determining that a load of the target virtual entity is empty, continuing to perform the scheduling operation before the time slice corresponding to the target virtual entity runs out (52 and 61. The system constantly monitors for idleness. Thus, prior art teaches “…before the time slice corresponding to the target virtual card runs out, continuing to performing the above scheduling operation in response to determining that the load of the target virtual card is empty” refers to a second interpretation of the claim language where it refers to the overall scheduling operation of repeated scheduling of next workload when the current workload one runs out of work (instead of the scheduling operation of claim 4 that schedules the target virtual card time slice, and thus can be interpreted as letting the said target virtual card time slice finish.)). Rationale to combine same as claim 6 above.
As for claim 10, Liu also teaches in response to determining that in the virtual card set there is a designated virtual card having a non-empty load and no corresponding time slice allocated, determining a target virtual card in the designated virtual card and allocating a corresponding time slice to the target virtual card (paragraphs 41-42); and continuing to perform the scheduling operation, in response to determining that the time slice corresponding to the target virtual card runs out (paragraph 44-45 in view of paragraph 47. Teaching letting the current time slice run out then repeat scheduling next workload)
In addition, Cheng teaches the same (paragraphs 52 and 61 in view of 34 teaching scheduling the target virtual card time slice early when current workload became idle (ran out of work), and paragraph 40 teaching letting time-slice for virtual entity run to expiration when it has work and non-idle). Rationale to combine same as claim 6 above.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KEVIN X LU whose telephone number is (571)270-1233. The examiner can normally be reached M-F 10am-6pm.
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, Lewis Bullock can be reached on 5712723759. 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.
/KEVIN X LU/Examiner, Art Unit 2199
/LEWIS A BULLOCK JR/Supervisory Patent Examiner, Art Unit 2199
Prosecution Insights