Abstract: Extending object-schema-based application programming interfaces (APIs) is described. According to one embodiment, a method generally includes receiving, from a user, a schema defining an extension to the API and a reference to a parent node in a graph projection of the API. An API system updates the graph projection of the API to include a node representing the extension and navigable path to the node representing the extension. The API system processes a request from the user by traversing through the updated graph projection of the API, the request representing the navigable path to the node representing the extension.

Abstract: An apparatus for performing a disruptive task based on a user state is disclosed. The apparatus may include a processor and a memory that stores code executable by the processor to identify a disruptive task to be performed at the apparatus. The processor obtains biometric data from a user of the apparatus. The processor performs the disruptive task in response to the biometric data indicating a predetermined user state. A method and computer program product also perform the functions of the apparatus.

Abstract: To estimate the relation between virtual machine performance and scheduling parameters which are set in a scheduler which allocates CPU time to a virtual machine, provided is a performance prediction device, comprising a measuring unit and an estimation unit. The measuring unit sets the scheduling parameters to sample values, operates a virtual computer system, and measures CPU allocation delay time from when the virtual machine requests an allocation of the CPU to when the CPU is allocated to the virtual machine and turnaround time (TAT) of the virtual machine. Based on the results of the measurements, the estimation unit generates an estimation formula with the scheduling parameters and the CPU allocation delay time as explanatory variables and an estimated TAT as the response variable.

Abstract: A method for executing a data processing pipeline may be provided. The method may include identifying a file providing a runtime environment required for executing a series of data processing operations comprising the data processing pipeline. The file may be identified based on one or more tags associated with the data processing pipeline. The one or more tags may specify at least one runtime requirement for the series of data processing operations. The file may be executed to generate an executable package that includes a plurality of components required for executing the series of data processing operations. The series of data processing operations included in the data processing pipeline may be executed by at least executing the executable package to provide the runtime environment required for executing the series of data processing operations. Related systems and articles of manufacture, including computer program products, are also provided.

Abstract: A workflow scheduling system includes a first processor configured to schedule a plurality of workflows each including a plurality of tasks; a plurality of second processors configured to form a predetermined number of logical computation units and execute the scheduled workflows in parallel; and a memory that stores information about a plurality of task groups each of which includes one or more tasks from one or more of the workflows. The first processor is configured to, based on the stored information, instruct the second processors to execute the scheduled workflows while limiting a total number of the workflows simultaneously executed by the second processors to the predetermined number for each of the task groups.

Abstract: Provided is a method and apparatus for implementing microkernel architecture of industrial server. The method includes calculation of dependency of control programs according to a microkernel task type weight and a microkernel task priority weight and/or a control program running time weight prior to startup of a system, and determination of the number of the control programs running on each physical core and each control program running on multiple physical cores according to the dependency.

Abstract: A simultaneous multi-threading (SMT) processor core capable of thread-based biasing with respect to execution resources. The SMT processor includes priority controller circuitry to determine a thread priority value for each of a plurality of threads to be executed by the SMT processor core and to generate a priority vector comprising the thread priority value of each of the plurality of threads. The SMT processor further includes thread selector circuitry to make execution cycle assignments of a pipeline by assigning to each of the plurality of threads a portion of the pipeline's execution cycles based on each thread's priority value in the priority vector. The thread selector circuitry is further to select, from the plurality of threads, tasks to be processed by the pipeline based on the execution cycle assignments.

Abstract: Examples disclosed herein relate to determining data usage effectiveness. A processor may determine data usage effectiveness information related to an entity's workflow based on workflow data collected related to the entity's operations and metrics determined based on the data. The determination may be based on a comparison of the information related to the entity's workflow to target workflow information and priority information associated with the target workflow information. The processor may output information related to the determined data usage effectiveness.

Abstract: Provided are a method, an apparatus, and an electronic device for request scheduling of hybrid edge computing, wherein the request scheduling method includes: obtaining situations of a topological structure, computing resources and communication resources of a communication network; determining the number of hybrid edge servers placed in the communication network; dividing a problem of an optimal request scheduling path of the hybrid edge server into sub-problems of Mixed-Integer Non-Linear Programming (MINLP) models according to the situations of the topological structure, the number, the computing resources and the communication resources of the network; establishing the MINLP model according to the sub-problems respectively; converting the MINLP model into a Mixed-Integer Linear Programming (MILP) equation; and solving the MILP equation to obtain a placement node of the hybrid edge server and a target server for processing communication requests.

Abstract: A queue-based task management system is provided. Unlike conventional queue-based task management systems, the system described herein does not handle all tasks in the conventional manner. Rather, tasks can be associated with one of several modes, including: a queue-mode (always storing the task in the queue) and an immediate-mode (record the task, but execute it immediately if possible). The two modes may be controlled programmatically to optimize utilization of system resources. Immediate-mode is implemented by monitoring available task system resources, and executing the task immediately in-process if there are resources available; otherwise the task is delegated to the queue. This in-process execution allows the sharing of parent task resources.

Abstract: Apparatuses, methods and storage medium associated with cluster computing are disclosed herein. In embodiments, a server of a computing cluster may include memory. input/output resources, and one or more processors to operate one of a plurality of application slaves of an application master; wherein the other application slaves are operated on other servers, which, together with the server, are members of the computing cluster. The server may further include a service assurance manager agent to manage allocation of the one or more processors, the memory and the input/output resources to the application slave, to assure compliance with a node level service level agreement, derived from an application level service level agreement, to contribute to proximate assurance of compliance with the application level service agreement; wherein the application level service agreement specifies the aggregate service level to be jointly provided by the application master and slaves.

Abstract: There is provided an information processing apparatus, a service processing apparatus, an information processing method, a program, and an information processing system, which are capable of appropriately determining a service processing apparatus as a connection destination from among a plurality of service processing apparatuses without using any special apparatus, the information processing apparatus including: a transmission control unit that causes a first communication unit to transmit an inquiry about whether or not processing of a first service requested by a user is able to be performed to a plurality of service processing apparatuses; and a connection destination determination unit that determines any service processing apparatus capable of performing the processing of the first service from among the plurality of service processing apparatuses as a connection destination on a basis of response information that is received from one or more of the plurality of service processing apparatuses.

Abstract: Various aspects are disclosed for optimization of dependent systems for serverless frameworks. In some examples, a load test executes instances of a function on a dependent system to generate datapoints. The datapoints are organized, using a clustering algorithm, into an acceptable group and at least one unacceptable group. A maximum number of concurrent instances of the function is determined based on a number of instances specified by at least one datapoint selected from the acceptable group. A live workload is performed on the dependent system. The live workload includes instances of the function that are assigned to the dependent system according to the maximum number of concurrent instances.

Abstract: The present invention relates to a communication system (1) comprising one or several execution node(s) (2) able to execute one or several microservice(s) (5), a computer device, called «host server» (3), including several routers (30) constituting an intermediate communication interface between each execution node (2) and the outside of the communication system (1), a heterogeneous computing platform (4), consisting of a set (40) of hardware and software or executable code for the access to and deployment of the microservices (5) on the system in a Java runtime environment (J) on the host server (3) and the execution nodes (2) allowing the execution of computer programs based on the Java language; the communication system (1) allows the creation of ephemeral microservices (5) by the use of a key/value system (6) stored in a distributed memory cache (8) at each creation by referencing each microservice (5) by filenames deposited in the system by a developer (10) and using an asynchronous TCP exchange protocol

Abstract: Exploiting FPGAs for acceleration may be performed by transforming concurrent programs. One example mode of operation may provide one or more of creating synchronous hardware accelerators from concurrent asynchronous programs at software level, by obtaining input as software instructions describing concurrent behavior via a model of communicating sequential processes (CSP) of message exchange between concurrent processes performed via channels, mapping, on a computing device, each of the concurrent processes to synchronous dataflow primitives, comprising at least one of join, fork, merge, steer, variable, and arbiter, producing a clocked digital logic description for upload to one or more field programmable gate array (FPGA) devices, performing primitive remapping of the output design for throughput, clock rate and resource usage via retiming, and creating an annotated graph of the input software description for debugging of concurrent code for the field FPGA devices.

Abstract: A method and system for invoking an event-based package module, in which the method is executed by a computer, and includes modulating a program package in which a plurality of functions related to a specific event has been implemented into an event module using a programming language based on reflection, loading the program package corresponding to a plurality of events, analyzing the loaded program package using the reflection, and structurizing the class and function of the analyzed program package for each event.

Abstract: Publishing events. A method includes a publisher sending a message comprising one or more events to a domain. The domain is a container comprising a group of domain topics. Each domain topic in the domain is relevant by subject matter to the one or more events by virtue of belonging to the container. The method further includes, at the domain determining the appropriate domain topics from the group of domain topics for each event in the one or more events. The method further includes identifying subscribers to the appropriate domain topics. The method further includes sending notifications to the subscribers of the domain topics of the one or more events.

Abstract: According to one set of embodiments, a computer system can expose, to a non-native software program running in a web browser of the computer system, one or more native OS APIs implemented by an OS of the computer system, where the exposing enables the non-native software program to invoke the one or more native OS APIs as part of its execution within a runtime environment provided by the web browser.

Abstract: Embodiments of the present invention provide an application interaction method, apparatus, and system. The method includes: receiving, by a transmission configuration module, a first instruction, where the first instruction includes an identifier of a first application and an identifier of a second application; determining, by the transmission configuration module, deployment information of the first application and deployment information of the second application based on the first instruction; determining, by the transmission configuration module, an information transmission mode between the first application and the second application based on the deployment information of the first application, the deployment information of the second application, and a transmission mode selection policy; and using, by the first application, the transmission mode to transmit information to the second application.

Abstract: Systems and methods for predictive technology incident reduction are disclosed. In one embodiment, in an information processing apparatus comprising at least one computer processor, a method for predictive technology incident reduction may include: (1) receiving a change record for a proposed change to a computer application or a computer network infrastructure; (2) analyzing the potential change for an adverse potential cross impact with another computer application or a computer system; (3) predicting a probability of failure and an impact of the proposed change using a model; (4) in response to a low predicted probability of failure, or a high predicted probability of failure with a low predicted impact: approving the proposed change; and implementing the proposed change; and (5) in response to a high predicted probability of failure and a high predicted impact, rejecting the proposed change.

Abstract: A system and method for automated computer system diagnosis and repair is provided. The system and method may allow for the automated diagnosis and repair of computer system errors without needing a user to manually engage a service desk, help desk, and/or the like. A user terminal may interact with a tech care module to generate a user repair request comprising the computer system error and data relating to the error. The system may analyze the user repair request and retrieve a repair script relating to the computer system error. The system may execute the repair script on the user terminal to resolve the computer system error. The system may also implement reporting modules to ensure that the computer system error is remediated and to update the repair scripts.

Abstract: A memory device includes a memory module and a control module. The control module is coupled to the memory module and is configured to store data into the memory module according to a first mapping table. The control module includes a storing unit and a guaranteeing unit. The storing unit is configured to store the first mapping table. The guaranteeing unit is coupled to the storing unit and is configured to determine whether the first mapping table is correct or not. The guaranteeing unit is further configured to issue an error signal in a state where the first mapping table is incorrect.

Abstract: One embodiment of the present invention sets forth a technique for testing microservices in distributed computing systems. The technique includes routing a first traffic sample associated with request traffic from one or more client devices to a first instance of a first service implemented in a distributed computing system, wherein the first instance generates a first plurality of request responses in response to a first plurality of data requests included in the first traffic sample and transmits the first plurality of request responses to a second service included in the sequence of services, and routing a second traffic sample associated with the request traffic to a second instance of the first service, wherein the second instance generates a second plurality of request responses in response to a second plurality of data requests included in the second traffic sample and transmits the second plurality of request responses to the second service.

Abstract: In some examples, a system for determining whether an operating system fault has occurred includes data storage and a processing system. The data storage may store image data indicative of a computing system display output. The processing system may access the stored image data. The processing system may determine that the computing system display output corresponds to a fault display output associated with a fault state of an operating system, which determination may include the processing system determining an extent of similarity between the accessed image data and a reference image associated with the fault display output. The processing system may generate a fault indication responsive to determining that the computing system display output corresponds to the fault display output.

Abstract: Disclosed embodiments relate to automatically providing updates to at least one vehicle. Operations may include receiving, at a server remote from the at least one vehicle, Electronic Control Unit (ECU) activity data from the at least one vehicle, the ECU activity data corresponding to actual operation of the ECU in the at least one vehicle; determining, at the server and based on the ECU activity data, a software vulnerability affecting the at least one vehicle, the software vulnerability being determined based on a deviation between the received ECU activity data and expected ECU activity data; identifying, at the server, an ECU software update based on the determined software vulnerability; and sending, from the server, a delta file configured to update software on the ECU with a software update corresponding to the identified ECU software update.

Abstract: A system, and corresponding method, is described for updating or calculating ECC where the transaction volume is significantly reduced from a read-modify-write to a write, which is more efficient and reduces demand on the data access bandwidth. The invention can be implemented in any chip, system, method, or HDL code that perform protection schemes and require ECC calculation, of any kind. Embodiments of the invention enable IPs that use different protections schemes to reduce power consumption and reduce bandwidth access to more efficiently communicate or exchange information.

Abstract: A memory system includes a memory device and a memory controller. The memory device outputs data in response to a read command. The memory device includes a first function circuit which performs a first operation based on data stored in the memory device, in response to the read command, to generate first processed data. The memory controller provides the read command to the memory device in response to a read request received from a host. The memory controller receives status information associated with performing the first operation. The memory controller includes a second function circuit which performs a second operation based on the first processed data to generate second processed data. A manner of the second operation varies based on the status information.

Abstract: Designs of nonvolatile memory systems and memory operations that enable allocating reliability values associated with the hard data values from memory cells and performing an error correction operation for the hard data based on the reliability values. One example of a memory system includes a nonvolatile memory device including a plurality of memory cells, and a controller suitable for obtaining hard data by performing a hard read operation based on a hard read voltage for the memory cells, and performing an error correction operation for the hard data based on reliability values.

Abstract: There is provided a method of obtaining one or more parity symbols (PS) of an encoding of information symbols (IS) according to a linear cyclic code, the method comprising: upon a permutation of information symbols (IS), generating data indicative of parity coefficients of a row of a generator matrix associated with the linear cyclic code, computing, for each given parity coefficient, a first data in accordance with, at least, the given parity coefficient and the first IS; updating, by the processing circuitry, for each given parity coefficient of the one or more parity coefficients, the first data, in accordance with, at least, the given parity coefficient and the respective IS; and upon meeting a parity completion criterion for a given parity coefficient, deriving a parity symbol from the respective first data, thereby obtaining the one or more parity symbols of the codeword of the linear cyclic code.

Abstract: A storage device includes a nonvolatile memory (NVM) device having a plurality of memory blocks and a control circuit configured to perform a read for copy-back operation in response to a receipt of a corresponding command. The control circuit performs the read for copy-back operation by reading page data from a source memory block of the plurality, generating a syndrome from the read page data, outputting the syndrome, receiving error location data in response to outputting the syndrome, correcting the read page data using the received error location data, and writing the corrected read page data to a target memory block among the plurality.

Abstract: A non-volatile (NV) memory accessing method using data protection with aid of look-ahead processing, and associated apparatus such as memory device, controller and encoding circuit thereof are provided. The NV memory accessing method may include: receiving a write command and data from a host device; obtaining at least one portion of data to be a plurality of messages, to generate a plurality of parity codes through look-ahead type encoding, wherein regarding a message: starting encoding a first partial message to generate a first encoded result; applying predetermined input response information to a second partial message to generate a second encoded result, and combining the first and the second encoded results to generate a first partial parity code; and starting encoding the message to generate a second partial parity code, and outputting the first and the second partial parity codes to generate a parity code; and writing into the NV memory.

Abstract: According to an embodiment, a memory controller obtains first data in a first page using a first voltage, obtains a first shift amount based on a first and second number. The first and second numbers represent numbers of bits each of which has different values in a first and second manner between the first data and first expected data. The controller obtains second data in the second page using a second voltage and a second shift amount, and obtains a third shift amount based on a third and fourth number, the third and fourth numbers respectively represent numbers of bits each of which has different values in the first and second manner between the second data and second expected data.

Abstract: Aspects of the present disclosure configure a system component, such as memory sub-system controller, to defer performance of an error-correction parity calculation for a block of a memory components of the memory subsystem. In particular, a memory sub-system controller of some embodiments can defer (e.g., delay) performance of an error-correction parity calculation and can defer the error-correction parity calculation such that it is performed at a time when the memory sub-system satisfies an idle state condition.

Abstract: Embodiments of the present application provide operation retry methods and apparatuses. One exemplary method includes: detecting whether an operation of a job fails; if there is an operation failure, collecting statistics on a progress of the job; calculating a retry interval according to the progress of the job; and re-executing the operation after waiting for the retry interval. When an operation of a job fails, the retry interval can be adaptively calculated according to the progress of the job. This can help increase the length of the retry interval, especially for a long job. The job can thereby be retried dynamically, to cope with longer service interruption. Embodiments of the present application can help avoid waste of resource waste caused by job failure and job re-execution, and reduce the retry costs while ensuring the success rate of the job.

Abstract: In one example, a method for a distributed model of data ingestion includes receiving multiple asynchronous backup requests from a plurality of clients, and performing primary processing with respect to data elements of a first one of the backup requests. Next, a job concerning a secondary process is added to a job queue. A handoff is then made for performance of the secondary processing with respect to data elements of the first backup request. The job is then accessed and the secondary processing is performed with respect to the data elements of the backup request, wherein the primary processing and secondary processing are respective first and second stages of an ingestion process, and the primary processing is performed prior to the secondary processing.

Abstract: The disclosed computer-implemented method for backing-up an eventually-consistent database in a production cluster may include (1) forming, on a production node, a stable copy of production data, (2) provisioning storage on a backup node based on an amount of data in the stable copy and a replication factor, (3) transferring information from the stable copy to a backup copy on the backup node, (4) performing record synthesis on the backup copy to merge record updates into complete backup records, (5) identifying and discarding any stale records and any redundant records in the complete backup records, and (6) transferring the complete backup records from the backup node to a cloud storage device. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: A data store manager of a multi-data-store storage system detects an error condition associated with applying a state change indicated in a particular journal entry of the storage system to a first data store. A first redaction entry inserted in the journal at some point after the particular entry is identified. The first redaction entry indicates the particular entry as a redaction target. After verifying that the journal does not include another redaction entry targeting the first redaction entry, processing of the particular entry is terminated and processing of later-inserted entries can proceed.

Abstract: A method for processing query fault, where a database server receives a query statement and generates a corresponding query plan tree including multiple layers of operators in a pipeline relationship, and each layer includes operation symbols having logical relationship with each other. The server executes the query statement according to the query plan tree, extracts intermediate status information of a faulty operator when a fault occurs in a process of executing the query statement, updates operation symbols of the faulty operator and a logical relationship among the operation symbols according to the query plan tree and the intermediate status information to obtain a reconstructed query plan tree, and continues to execute the query statement according to the reconstructed query plan tree after the fault is recovered.

Abstract: A method of error correction in a nonvolatile flash memory of NAND type, the memory including a first and a second, identical, copies of software to be loaded, a page of a block occupied by data of the first copy of the software being corrupted, and a page of a block occupied by data of the second copy of the software being corrupted, the method including: the replacement of the block occupied by the second copy of the software and including a corrupted page with the identical block occupied by the first copy of the software, making it possible to correct the second copy of the software; the re-execution of the boot loading program; the loading of the second copy of the software.

Abstract: In various embodiments, the system includes augmented database drivers that are configured to automatically retry execution of write operations if a failure is encountered. In some embodiments, a database daemon is configured to manage the database functionality for a respective database node (e.g., primary or secondary node). Retrying execution of write operations allows the drivers to automatically retry certain write operations a threshold number of times if, for example, a network error is encountered, or if a healthy primary node is not available.

Abstract: A method for recovering content stored in virtual machines is provided. This method includes isolating a data object to be recovered from a crash or error from incoming and outstanding processes; backing out uncommitted changes of the data object based on undo logs; determining a most recent checkpoint for the data object; reinitializing the data object from the most recent checkpoint to produce a reinitialized data object; and populating the reinitialized data object to produce a recovered data object by replaying, from a write-ahead log.

Abstract: Each server device of a plurality of server devices may perform a method that includes detecting, by a second server device, that a first storage access virtual machine (SAVM) operating on a first server device has failed, the first SAVM managing access to a first data store of a shared file system by a first client operating on the first server device using a first storage address. The method also includes directing a second SAVM operating on the second server device to manage access to the first data store and granting the second SAVM access to a first state information file stored within the shared file system, the first state information file being associated with the first storage address. The method can also include restoring to the first SAVM, when it becomes operational, management of access to the first data store and access to the first state information file.

Abstract: Example implementations described herein are directed to a storage descriptor data structure that can represent characteristics of the dataset stored in an underlying volume without requiring the dataset in the volume to be migrated for processing. Such data structures are more compressed than the original data structure and can therefore facilitate storage allocation before the migration is conducted.

Abstract: Disclosed are hardware and techniques for building runbooks for new computer-implemented processes by correlating break events from the new processes with break events extant in existing runbooks for existing computer-implemented processes. In addition, fix events associated with the correlated break events are evaluated to determine the likelihood that they will be able to fix the error condition which caused the break event from the new process. The fix events are presented to a human operator who may select and test each fix event to determine if the error condition is directed and, if so, the correlated break event associated with the fix event are merged together and added to a new runbook for the new computer-implement process.

Abstract: A display device includes a display panel including a plurality of data lines; a plurality of data drive ICs configured to provide data voltages to the data lines; and a controller configured to rearrange video data received from an external device and transmit region-wise video data to the plurality of data drive ICs, to receive coordinates of information regions included in the video data and store video data of the information regions in advance, to calculate a checksum for checking an error in the region-wise video data based on the coordinates, to transmit the coordinates and the checksum to the data drive ICs and receive feedbacks about comparison results with respect to the checksum, and when a data drive IC that has failed is confirmed based on the feedback results, to update video data of the information region displayed by the data drive IC to region-wise video data of a normal data drive IC and output the video data.

Abstract: A system includes a distributed data storage system disseminated across worker machines connected by a network. A distributed data storage management module has instructions executed by a processor to utilize data block identifiers to track data block accesses to the distributed data storage system. A sampling module with instructions executed by the processor receives a new sample request from a client machine connected to the network. Initial data block samples are gathered from the distributed data storage system during a first time period. A revised sample request is received from the client machine during the first time period. The initial data block samples are gathered. New data block samples are collected from the distributed data storage system. The initial data block samples and the new data block samples are combined to form cumulative data block sample results. The cumulative data block sample results are supplied to the client machine.

Abstract: A storage apparatus includes a controller unit and output unit. The controller receives cyclic information that is cycled within a specific group of storage apparatuses with properties close to a relevant storage apparatus. The cyclic information includes performance information of each storage apparatus in the specific group and evaluation information indicative of an evaluation result of the performance information of each storage apparatus. The controller acquires the performance information including the evaluation results, combines it with the cyclic information, and transmits the cyclic information to a storage apparatus which is a next cycling destination. A similarity level may be calculated based on configuration information and a predetermined weight. A specific group may be formed by checking whether storage apparatuses are capable of communicating with each other, so a cyclic path can be formed. The output unit may transmit performance information as information to be displayed on a user terminal.

Abstract: Systems and methods are described for the collection and transmission of virtual machine resource operations information. Individual agents on virtual machine resources collect and store operations information in accordance with a current operations information collection configuration. The individual agents will initiate a transmission of the collected operations information, including an identification of the current operations information collection configuration being implemented by the respective agent. Responsive to the receipt of the transmission of the collected operations information, the monitoring processing service makes a determination of whether the current operations information collection configuration is appropriate or should otherwise be maintained or changed. The monitoring processing service will include the identification of the new operations information collection configuration in the message responsive to the transmission of the collected operations information from the agent.

Abstract: A software instruction code repair system comprising an instruction code example pool. The example pool comprises a set of good instruction code examples and a set of bad instruction code examples. The software instruction code repair system further comprises a sequence-to-sequence (seq2seq) network that is configured to generate a corrected instruction code example, based on one example of the set of bad instruction code examples. The software instruction code repair system further comprises a discriminator configured to randomly select one of the corrected instruction code example and one instruction code example of the set of good instruction code examples to produce a selected instruction code example. The discriminator is further configured to make a determination that the selected instruction code example is most likely taken either the instruction code example pool or the seq2seq network.

Abstract: The invention relates to the verification of properties of an interactive component having a recursive tree structure of subcomponents. The verification comprises creating an oriented graph from the tree, said oriented graph comprising arcs for the child relationships of the tree, and arcs for couplings creating upon couplings between the components. The verification comprises creating a second graph representative of the property, and verifying if the second graph is exhibited within the first oriented graph.