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 .
DETAILED ACTION
This action is in response to application filed 12/31/2024.
Claims 1-20 are pending in this application.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 03/28/2025, 12/29/2025 has been placed in record and considered by the examiner.
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 of this title, 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 set forth in Graham v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied 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.
Claims 1-3, 6, 8-15, 17, 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Ambichl et al. (US 2020/0042426 A1) in view of Banerjee et al. (US 2026/0037224 A1) in further view of Safavi (US 2024/0430153 A1).
Regarding claim 1, Ambichl discloses a computing system comprising:
storage media; and processing circuitry in communication with the storage media (fig. 1: Monitoring server 100 with multiple processors and repositories. [0051]: A monitoring server 100 interacts with a heterogeneous set of agents and APIs 102 deployed to a monitored environment 101), the processing circuitry configured to:
receive time series data comprising performance indicators for elements of a plurality of layers of a system having host computing devices that execute workloads for an application, the time series data generated by the elements ([0056]: scalar measurement records 104 containing measurement data describing components of the monitored environment, like resource usage measures, transaction records 105 containing data describing transactions executed by components of the monitored environment, topology records 106, describing elements of the monitored environments and their vertical (e.g. processes executed by host computing system) and horizontal (e.g. communication activities between monitored processes) relationships. [0059]: Scalar measurement records 104 are received by a scalar measurement processor 110 and analyzed to create measurement time series. The time series data may be stored in a measurement repository. [0066]: The service extractor 116 may also extract measurement data corresponding to identified services from transaction trace data and create service measurement data 115 which may be stored in the measurement repository in form of time series. Service measurement data 115 also contains topology identification data that identifies the service topology entity related to the service measurement data);
create, using the time series data ([0168]: A time series record 560 may contain but is not limited to topology identification data 561 identifying the topology entity that is the origin of the time series data stored in the time series record), a dependency graph comprising nodes and edges ([0005]: the proposed monitoring system uses the topology data and a set of causality rules to estimate causal relationships between pairs of identified abnormal operating conditions to create directed graphs that describe the causal dependencies between multiple abnormal operation conditions. [0022]: Local causality graphs that share equivalent events may be merged. Afterwards, newly merged causality graphs may be compared with causality graphs that were already provided to users of the monitoring system to identify equivalent events with those causality graphs. [0071]: Local causal dependency graphs may be represented by a set of causality event nodes 720 connected by causality graph edges 730);
determine, based on the time series data, an anomaly in the performance of a workload of the workloads for the application ([0067]: A trigger event generator 121 may cyclically fetch selected measurement time series (e.g. time series describing services used by transactions that provide value for end-users of the application) and use an anomaly detector 122 to determine whether those measures indicate abnormal operating conditions. [0069]: Trigger events may contain data specifying the type and the extend of an abnormal operation condition, the location of the abnormal operation condition and data describing the temporal extend of the abnormal operation condition. Causality graph event nodes 720 as described in FIG. 7 may be used to represent trigger events);
determine, based on a dependency of the workload on an element of the network layer within the dependency graph (fig. 13, [0241]: determine the root cause event within a previously calculated causality event graph), an issue with the element of the network layer as a root cause of the anomaly in the performance of the workload ([0018]: recursive, topology model driven search for other abnormal operation conditions for which the probability of a causal relationship with the trigger event exceeds a certain threshold. The causality search may start at topology element on which the trigger event was executed. [0053]: Identified transaction execution related anomalies are located on a topology element, and the connection data of the topology model may be used to identify topology entities that are related to the location); and
However, Ambichl does not disclose create, based on mapping, Graphics Processing Units (GPUs) of a compute layer of the plurality of layers to network devices of a network layer of the plurality of layers, a dependency graph, the nodes representing the elements of the plurality of layers and the edges representing cross-layer relationships between pairs of the elements, each of the elements of each pair being of different layers of the plurality of layers.
In an analogous art, Banerjee discloses create, based on mapping, Graphics Processing Units (GPUs) of a compute layer of the plurality of layers to network devices of a network layer of the plurality of layers, a dependency graph ([0011]: The fat tree configured network also includes a collective of processing units, e.g., graphics processing units (GPUs) that communicate with one another via the tiers of nodes. [0029]: the network 100 has a fat tree topology, configured to provide in network compute (INC). The network 100 includes multiple processing units 102 a-102n (referred to collectively as processing units 102). The processing units 102 (e.g. compute layer) may be in the form of graphics processing units (GPUs), each having their own (e.g. mapping) network interface controller (NIC) 104 (e.g. network layer). The network 100 also includes multiple first switches 106 a-106 n (referred to collectively as first switches 106) in a first tier 108 and multiple final switches 110 a-110 n (referred to collectively as final switches 110) in a final tier 112 (e.g. network devices of a network layer. [0050]: determining, the INC manager selects a first switch as a root, wherein the first switch is included within a tier of switches having intermediate tiers of switches located between the tier and the plurality of processing units within the fat tree configured network), the nodes representing the elements of the plurality of layers and the edges representing cross-layer relationships between pairs of the elements, each of the elements of each pair being of different layers of the plurality of layers ([0034]: there are generally more processing units 202, although in configurations, there may be fewer processing units 202. The example arrangement 200 also includes a first tier Tier-0, a second tier Tier-1, and a final tier Tier-2. Tier-0 includes four nodes (e.g., switches) 204 a-204 d (referred to collectively as nodes 204. [0042]: The tree T may be used to execute collective computing operations or workloads, e.g., AI/ML collective computing operations. More particularly, ALUs of the nodes of the trees, e.g., switches of the trees, may be used to more efficiently and/or quickly execute collective computing operations or workloads, e.g., AI/ML collective computing operations, in conjunction with the processing units 202.).
Therefore, it would have been obvious before the effective filed date of the claimed invention to a person having ordinary skill in the art to modify Ambichl to comprise “create, based on mapping, Graphics Processing Units (GPUs) of a compute layer of the plurality of layers to network devices of a network layer of the plurality of layers, a dependency graph, the nodes representing the elements of the plurality of layers and the edges representing cross-layer relationships between pairs of the elements, each of the elements of each pair being of different layers of the plurality of layers” taught by Banerjee.
One of ordinary skilled in the art would have been motivated because it would have enabled to represent the relationships between processing units and network devices and to model the communication topology in which AI/ML workloads are executed in order to support more efficient execution of collective AI/ML computing operations or workloads (Banerjee, [0042], [0054]).
However, Ambichl-Banerjee does not disclose perform a remedial action for the issue with the element of the network layer as a root cause of the anomaly in the performance of the workload.
In an analogous art, Safavi discloses perform a remedial action for the issue with the element of the network layer as a root cause of the anomaly in the performance of the workload ([0004]: identifying the root cause of the issue and invoking remedial actions in a timely, efficient and cost-effective manner. [0007]: determining, from the network data, a time series of statistics; aggregating, at the computing device, the time series of statistics from the plurality of network devices to produce aggregated statistics; evaluating, at the computing device, the aggregated statistics to determine an anomaly in the network, the anomaly having an associated anomaly time period).
Therefore, it would have been obvious before the effective filed date of the claimed invention to a person having ordinary skill in the art to modify Ambichl-Banerjee to comprise “perform a remedial action for the issue with the element of the network layer as a root cause of the anomaly in the performance of the workload” taught by Safavi.
One of ordinary skilled in the art would have been motivated because it would have enabled detecting congestion and other network problems, thereby identifying the root cause of the issue and invoking remedial actions in a timely, efficient and cost-effective manner (Safavi, [0004).
Regarding claim 2, Ambichl-Banerjee-Safavi discloses the computing system of claim 1, wherein the plurality of layers include an application layer, the compute layer, and the network layer (Ambichl, [0014], [0027]: The topology-related data contains data describing virtualization infrastructure used by the monitored system, like virtualization management components and the virtualized host computing systems provided by those virtualization management components, host computing systems including virtualized and non-virtualized host computing systems, the processes executed on those host computing systems (e.g. compute layer) and the services (e.g. application layer) provided by those hosts. [0070]: The analysis performed by causality estimator also uses the topology model 118 and the topology coordinates of available monitoring data to identify topological entities that are connected (e.g. network layer) (either via communication activities, like processes communicating using a computer network or via shared resources).
Regarding claim 3, Ambichl-Banerjee-Safavi discloses the computing system of claim 2, wherein the time series data comprises one or more of: application-specific metrics for the workloads for the application of the application layer; one or more logs for a collective communication library (CCL) of the application layer; node level telemetry associated with the plurality of host computing devices of the compute layer; Graphics Processing Unit (GPU) telemetry associated with a plurality of GPUs of the plurality of host computing devices of the compute layer; Remote Direct Memory Access (RDMA) over Converged Ethernet (RoCE) information of the compute layer; or flow telemetry or network telemetry of the network layer (Ambichl, [0004], [0066]: extract measurement data corresponding to identified services from transaction trace data and create service measurement data 115 which may be stored in the measurement repository in form of time series).
Regarding claim 6, Ambichl-Banerjee-Safavi discloses the computing system of claim 1. Ambichl disclose wherein, to create the dependency graph, the processing circuitry is further configured to create, based on telemetry of the time series data (Ambichl, [0168]: A time series record 560 may contain but is not limited to topology identification data 561 identifying the topology entity that is the origin of the time series data stored in the time series record), a first edge between the workload of an application layer of the plurality of layers ([0005]: the proposed monitoring system uses the topology data and a set of causality rules to estimate causal relationships between pairs of identified abnormal operating conditions to create directed graphs that describe the causal dependencies between multiple abnormal operation conditions. [0022]: Local causality graphs that share equivalent events may be merged. [0071]: Local causal dependency graphs may be represented by a set of causality event nodes 720 connected by causality graph edges 730).
However, Ambichl does not disclose to create, based on GPU telemetry data, a first edge between the workload of an application layer of the plurality of layers and a GPU of the GPUs of the compute layer.
In an analogous art, Banerjee discloses to create, based on GPU telemetry data, a first edge between the workload of an application layer of the plurality of layers and a GPU of the GPUs of the compute layer ([0029]: the network 100 has a fat tree topology, configured to provide in network compute (INC). The network 100 includes multiple processing units 102 a-102n. The processing units 102 (e.g. compute layer) may be in the form of graphics processing units (GPUs), each having their own (e.g. mapping) network interface controller (NIC) 104 (e.g. network layer). The network 100 also includes multiple first switches 106 a-106 n (referred to collectively as first switches 106) in a first tier 108 and multiple final switches 110 a-110 n (referred to collectively as final switches 110) in a final tier 112 (e.g. network devices of a network layer). [0031]: The INC manager 116 is includes an algorithm 118 to create trees within the network 100 that provides routes within the network 100 through the various switches so that each of the processing units 102 is accessible via the first switches 104 and final switches 110 (and any intermediate switches). [0054]: the plurality of processing units and the switches execute a collective computing operation. For example, the tree T may be used to execute workloads, e.g., AI/ML workloads. ALUs of the nodes of the trees, e.g., switches of the trees, may be used to more efficiently and/or quickly execute collective computing operations or workloads).
Therefore, it would have been obvious before the effective filed date of the claimed invention to a person having ordinary skill in the art to modify Ambichl to comprise “to create, based on GPU telemetry data, a first edge between the workload of an application layer of the plurality of layers and a GPU of the GPUs of the compute layer” taught by Banerjee.
One of ordinary skilled in the art would have been motivated because it would have enabled to represent the relationships between processing units and network devices and to model the communication topology in which AI/ML workloads are executed in order to support more efficient execution of collective AI/ML computing operations or workloads (Banerjee, [0042], [0054]).
Regarding claim 8, Ambichl-Banerjee-Safavi discloses the computing system of claim 1, wherein the issue with the element of the elements residing at the plurality of layers comprises at least one of: insufficient Graphics Processing Unit (GPU) resources of a GPU of the GPUs of the compute layer; a temperature of the GPU of the GPUs of the compute layer; insufficient resources of a host computing device of host computing devices of the compute layer; congestion control for a Remote Direct Memory Access (RDMA) operation of the network layer; or network congestion of a network device of the network layer (Safavi, [0124]: The time series of statistics may be used to determine whether the network is experiencing a congestion issue or an anomaly and if so, determine the root cause of the congestion or anomaly. Once the root cause is identified, an automated and/or manual corrective action may take place). The same rationale applies as in claim 1.
Regarding claim 9, Ambichl-Banerjee-Safavi discloses the computing system of claim 1, wherein the processing circuitry is configured to detect the anomaly in the performance of the workload based at least in part on a rate of operations per iteration of the workload (Ambichl, [0067]-[0068]: A trigger event generator 121 may cyclically fetch selected measurement time series (e.g. time series describing services used by transactions that provide value for end-users of the application) and use an anomaly detector 122 to determine whether those measures indicate abnormal operating conditions. The trigger event generator may monitor the availability of selected topology entities, like host computer systems or processes that are used by a high number of transactions and are therefore of high importance for the functionality of the services provided by the monitored environment).
Regarding claim 10, Ambichl-Banerjee-Safavi discloses the computing system of claim 1, wherein the processing circuitry is further configured to output, based on the time series data and the dependency graph, an indication of a health of the elements residing at the plurality of layers (Ambichl, [0027]: vertical topology relationships like relationships between services and processes providing those services or processes and host computer systems executing those processes to identify causal dependencies. [0067]: A trigger event generator 121 may cyclically fetch selected measurement time series and use an anomaly detector 122 to determine whether those measures indicate abnormal operating conditions. The anomaly detector may use techniques like static thresholds or adaptive baselines (e.g. indication of a health) to determine the existence of abnormal operating conditions).
Regarding claim 11, Ambichl-Banerjee-Safavi discloses the computing system of claim 1, wherein the processing circuitry is further configured to output, based on the time series data and the dependency graph, a representation of an end-to- end path of communication between GPUs of two host computing devices of the host computing devices of the compute layer that execute the workload (Banerjee, [0046], [0054]: in configurations, the algorithm 118 may begin with gathering information from the nodes 204, 206, and 208. FIG. 2A includes a simple array of the processing units 202 that intend to be part of a collective of the network, e.g., the fat tree configured network. The tree T may be used to execute workloads, e.g., AI/ML workloads. ALUs of the nodes of the trees, e.g., switches of the trees, may be used to more efficiently and/or quickly execute collective computing operations or workloads, e.g., AI/ML collective computing operations, in conjunction with the processing units). The same rationale applies as in claim 1.
Regarding claim 12, Ambichl-Banerjee-Safavi discloses the computing system of claim 1, wherein the processing circuitry is further configured to: perform a remedial action for the root cause of the anomaly in the performance of the workload, wherein the remedial action comprises one of: configuring an end-to-end path of communication between GPUs of two host computing devices of the host computing devices of the compute layer that execute the workload; or configuring a congestion control scheme of a network device within the network layer that carries network traffic associated with the workload (Safavi, [0004]: detecting congestion and other network problems, identifying the root cause of the issue and invoking remedial actions in a timely, efficient and cost-effective manner. [0040]: mitigation module 135 may detect congestion and/or network anomalies as soon as they happen, identify the root cause of the anomaly, and invoke remedial actions. Remedial actions may involve automatic restarts of network devices, and changes in configuration. In one example, NMS 130 perform one or more of changing the configuration of a network device, changing the software version of the network device, or restarting the network device or a component of the network device (e.g. congestion control). In this way, NMS 130 using network anomaly detection and mitigation module 135 may improve network performance automatically). The same rationale applies as in claim 1.
Regarding claim 13, Ambichl-Banerjee-Safavi discloses the computing system of claim 1, wherein the workloads comprises one of: machine learning workloads; or graphics rendering workloads (Banerjee, [0054]: the plurality of processing units and the switches execute a collective computing operation. For example, the tree T may be used to execute workloads, e.g., AI/ML workloads. ALUs of the nodes of the trees, e.g., switches of the trees, may be used to more efficiently and/or quickly execute collective computing operations or workloads). The same rationale applies as in claim 1.
Regarding claims 14 and 20; the claims are interpreted and rejected for the same reason as set forth in claim 1.
Regarding claim 15; the claim is interpreted and rejected for the same reason as set forth in claim 2.
Regarding claim 17; the claim is interpreted and rejected for the same reason as set forth in claim 6
Regarding claim 19; the claim is interpreted and rejected for the same reason as set forth in claim 12.
Claims 4-5, 16 are rejected under 35 U.S.C. 103 as being unpatentable over Ambichl in view of Banerjee in view of Safavi, as applied to claim 3, in further view of Jo et al. (US 2024/0411562 A1)).
Regarding claim 4, Ambichl-Banerjee-Safavi discloses the computing system of claim 3. Ambichl discloses wherein the time series data comprises the one or more logs (Ambichl, [0132]: Log entries 334 may contain but are not limited to a timestamp 335 determining the time at which the logged event occurred and a log message 336 that textual and semi-structured describes the logged event).
However, Ambichl-Banerjee-Safavi does not disclose the one or more logs for the CCL of the application layer, the one or more logs comprising at least one of: a number of bytes transferred for the workloads by each host computing device of the host computing devices; or a number of operations performed for the workloads by each host computing device of the host computing devices.
In an analogous art, Jo discloses the one or more logs for the CCL of the application layer, the one or more logs ([0153]-[0155]: The modeling framework 820 may include an MPI (e.g. CCL) workload configurator 821, a metric model builder 822, and a workload profiler 823. The workload profiler 823 may receive the application and the job description on the application executable by the frontend 810 and perform profiling from various viewpoints. [0163]: The communication profiler 933 may receive an application and a job description on the application from the frontend 810. The communication profiler 933 may extract an MPI call trace with an argument while the application is executing. The extracted information may be stored in the profile database 934 (e.g. log)) comprising at least one of: a number of bytes transferred for the workloads by each host computing device of the host computing devices; or a number of operations performed for the workloads by each host computing device of the host computing devices (table 2, [0096]: The amount of communication (e.g. bytes transferred) in the MPI P2P communication may be determined based on a function…the MPI_Send function and/or the MPI_Recv function, the amount of communication may be determined based on the count argument indicating the data count and the datatype argument indicating the type of data among the arguments of the function. The size of data may be determined according to the value of the datatype argument, for example, a case in which the datatype argument is the MPI_INT may have a 4-byte size and a case in which the datatype argument is the MPI_DOUBLE may have an 8-byte size. The amount of communication may be determined based on the multiplication of the determined data size and the value of the count argument).
Therefore, it would have been obvious before the effective filed date of the claimed invention to a person having ordinary skill in the art to modify Ambichl-Banerjee-Safavi to comprise “the one or more logs for the CCL of the application layer, the one or more logs comprising at least one of: a number of bytes transferred for the workloads by each host computing device of the host computing devices; or a number of operations performed for the workloads by each host computing device of the host computing devices” taught by Jo.
One of ordinary skilled in the art would have been motivated because it would have determination of the amount of communication from MPI parameters for predicting communication latency and application execution time (Jo, [0096], [0106]).
Regarding claim 5, Ambichl-Banerjee-Safavi discloses the computing system of claim 1, wherein, to create the dependency graph, based on time series data, first edge between the workload of an application layer of the plurality of layers and one of a Network Interface Controller (NIC) or a switch of the network layer (Banerjee, [0011]: The fat tree configured network also includes a collective of processing units, e.g., graphics processing units (GPUs) that communicate with one another via the tiers of nodes. At each tier, and at each node of the tier, the reachability and bandwidth availability of the node from a lower tier is checked).
However, Ambichl-Banerjee-Safavi does not disclose to create, based on one or more logs for a collective communication library (CCL), a first edge between the workload of an application layer of the plurality of layers and network layer.
In an analogous art, Jo discloses to create, based on one or more logs for a collective communication library (CCL), a first edge between the workload of an application layer of the plurality of layers and network layer ([0163]: The communication profiler 933 may receive an application and a job description on the application from the frontend 810. The communication profiler 933 may extract an MPI call trace with an argument while the application is executing. The extracted information may be stored in the profile database 934).
Therefore, it would have been obvious before the effective filed date of the claimed invention to a person having ordinary skill in the art to modify Ambichl-Banerjee-Safavi to comprise “to create, based on one or more logs for a collective communication library (CCL), a first edge between the workload of an application layer of the plurality of layers and network layer” taught by Jo.
One of ordinary skilled in the art would have been motivated because it would have determination of the amount of communication from MPI parameters for predicting communication latency and application execution time (Jo, [0096], [0106]).
Regarding claim 16; the claim is interpreted and rejected for the same reason as set forth in claim 5.
Claims 7 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Ambichl in view of Banerjee in view of Safavi, as applied to claim 1, in further view of Raghavendra et al. (US 2018/0367414 A1)).
Regarding claim 7, Ambichl-Banerjee-Safavi discloses the computing system of claim 1, wherein, to create the dependency graph, the processing circuitry is further configured to create, a first edge between a GPU of the GPUs of the compute layer and one of a Network Interface Controller (NIC) or a switch of the network layer (Banerjee, [0018]: The fat tree configured network also includes a collective of processing units, e.g., graphics processing units (GPUs) that communicate with one another via the tiers of nodes. At each tier, and at each node of the tier, the reachability and bandwidth availability of the node from a lower tier is checked).
However, Ambichl-Banerjee-Safavi does not disclose to create, based on flow-level telemetry of the time series data, a first edge.
In an analogous art, Raghavendra discloses to create, based on flow-level telemetry of the time series data, a first edge ([0037]: network device state information for internetworking devices in the network and network traffic flow information for data packet flows through the network. The telemetry information is transformed into a temporal graph that is digitally stored in computer memory. [0064]: Each vertex of the temporal graph corresponds to a network device. Each edge of the temporal graph corresponds to a flow, or one or more portions of a flow, between two network devices).
Therefore, it would have been obvious before the effective filed date of the claimed invention to a person having ordinary skill in the art to modify Ambichl-Banerjee-Safavi to comprise “to create, based on flow-level telemetry of the time series data, a first edge” taught by Raghavendra.
One of ordinary skilled in the art would have been motivated because it would have enable to aggregate network telemetry information converted into temporal graph data in order to be used for network analytics, such as network assurance, network monitoring, link cost analysis, and network data flow analytics (Raghavendra, [0157], [0158]).
Regarding claim 18; the claim is interpreted and rejected for the same reason as set forth in claim 7.
Additional References
The prior art made of record and not relied upon is considered pertinent to applicants disclosure.
So et al., US 2025/0385828 A1: AI-Based Root Cause Analysis for Telecommunications Systema..
Brar et al., US 2024/0152409 A1Routing in a GPU Super-Cluster.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JUAN C TURRIATE GASTULO whose telephone number is (571)272-6707. The examiner can normally be reached Monday - Friday 8 am-4 pm.
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, Glenton B Burgess can be reached at (571)272-3949. 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.
/J.C.T/Examiner, Art Unit 2454
/DOUGLAS B BLAIR/Primary Examiner, Art Unit 2454
Prosecution Insights