Fault-tolerant computer system data comparing method serving operating system core layer

The invention relates to a fault-tolerant computer system data comparing method serving an operating system core layer. The data comparing service is provided for a double-module redundancy process in a fault-tolerant computer system by starting a core daemon process in a Linux operating system and executing data comparator logics. An event list is added in a core to be used as a message passage;the redundancy process and a data comparator are operated in a production-consumption way; the redundancy process packages the data to be written into a message packet to be inserted into a message list; and the comparator takes the message packet from the message list, analyzes the message packet according to a definition format, compares the data to be written of the redundancy process and finally returns a result to the redundancy process. The invention is realized in the operating system core layer, has no customized hardware, simple realization and good generality and is suitable for a process-grade ...

A multi-availability mechanism coexistence framework of concurrent storage system

This invention discloses a coexistence structure of a multi-use unit of a parallel storage system including a state test and control frame, a data service frame, a metadata service frame, a data synchronous frame, a customer end frame, a system management frame and a high usability module supporting load and unload of on-line modules, in which, the high usability module realizes that all frames call necessary interface functions in the mode of plug-in components and the frames call corresponding interfaces realized in the usable module based on the high usability kind used by logic data to finish specific function, and users can select the most suitable unit in the high usability unit to secure reliability and usability of logic data.

A concurrent storage system construction method for convenient expansion of storage node quantity

This invention discloses a structuring method for a parallel storage system convenient for expanding numbers of stored nodes, which uses a data distribution mode of selecting in order to structure a storage system, when stored nodes are increased, it' s only necessary to move data from the original stored node to newly increased nodes, and it' s not necessary to move data among the original storednodes to secure balance of the volume of stored nodes and consistency of the sequence selection mode.

Parallel implementation method and system of quantum attack resistant XMSS signature method for GPU (Graphics Processing Unit)

The invention discloses a parallel implementation method and system of an anti-quantum-attack XMSS signature method for a GPU (Graphics Processing Unit), and the parallel efficient implementation of the anti-quantum-attack XMSS signature method for the GPU is realized by combining a decision of a multi-level parallel scheme through a two-layer parallel method which comprises the parallelism on a first-layer structure, the parallelism of a second-layer Winternitz one-time signature related structure and the parallelism of L tree construction. By means of a two-layer parallel method including parallelism on a first-layer structure, parallelism of a second-layer related structure and parallelism of L-tree construction, efficient parallel execution of the XMSS algorithm on the GPU is achieved in combination with decisions of a multi-level parallel scheme.

Rootkit detection mechanism and detection method based on kernel-based virtual machine

The invention provides a rootkit detection method based on a kernel-based virtual machine. By adopting the rootkit detection method, the types of objects attacked by rootkit are subjected to abstract classification into static code segment, static data segment, dynamic distribution function and heap space data, and different protection and detection mechanisms are provided specific to different types; while specific rootkit sensitive information is obtained through a runtime module operating in a client machine, information interaction between the kernel-based virtual machine and the client machine is realized by adding a semantic processing pathway, and by extending a corresponding page-fault exception handler and a vmcall active trap-in mechanism, the normal execution of the client machine is guaranteed under the rootkit detection mechanism. The rootkit detection mechanism and detection method, provided by the invention, can be used for effectively preventing the attack specific to static ...

Method for disposing micro-OS with self-adaption transmission model

The invention discloses a deployment method for a micro-operation system with a self-adaptive transmission model. With the method, the transmission model of the micro-operation system is adaptively constructed through comparing a node scale N in a cluster and the size of the node number M synchronously deployed by a TFTP server at most. When N is smaller than or equal to M, the micro-operation system is deployed by adopting a C/S transmission model; when M is smaller than N, and N is smaller than or equal to 2M, the nodes are deployed in batches by adopting the C/S transmission model, and thebatching method overcomes the disadvantage that the deployment of the micro-operation system is failed caused by excessively heavy load of a TFTP server side; when N is larger than 2M, the C/S transmission model is automatically evolved into a multi-way tree transmission model, and by converting the nodes partially loaded with the micro-operation system into nodes which can provide TFTP service, the ...

Isomorphic calculation task grouping method used for eliminating memory access congestion

The invention provides an isomorphic calculation task grouping method used for eliminating memory access congestion. According to the method, grouping is conducted according to parallel income of isomorphic calculation task execution, and by enabling calculation regions and memory access regions of different groups to overlap, memory access congestion, caused when a many-core processor system runs, of applications is reduced or eliminated. According to the method, whether memory access congestion occurs or not does not need to be analyzed at the algorithm level, and a united and effective method is provided for memory access performance analysis and optimization of the applications with the isomorphic calculation task characteristics.

A method for constructing multi-availability mechanism coexistence framework of concurrent storage system

This invention discloses a coexistence structure of a multi-use unit of a parallel storage system including a state test and control frame, a data service frame, a metadata service frame, a data synchronous frame, a customer end frame, a system management frame and a high usability module supporting load and unload of on-line modules, in which, the high usability module realizes that all frames call necessary interface functions in the mode of plug-in components and the frames call corresponding interfaces realized in the usable module based on the high usability kind used by logic data to finish specific function, and users can select the most suitable unit in the high usability unit to secure reliability and usability of logic data.

Single key pair data parallel method and system for GPU (Graphics Processing Unit), medium and equipment

The invention discloses a single key pair data parallel method and system for a GPU (Graphics Processing Unit), a medium and equipment, which are characterized in that an XMSS tree construction process is reused during XMSS signature, Winternitz one-time signature is carried out, and thread count load balancing of XMSS tree construction is carried out; an authentication path and a root node are efficiently selected in parallel construction of an XMSS tree, and a parallel execution method of a highly configurable WOTS + signature is provided, so that efficient single-key parallel signature of an XMSS algorithm to data is realized; by deploying task distribution and a multi-stream scheme, efficient single-key parallel signature verification of the XMSS algorithm on data is realized.

Application program thread count dynamic regulating method used under isomerous many-core system

The invention discloses an application program thread count dynamic regulating method used under an isomerous many-core system. The application program thread count dynamic regulating method comprises the following steps: based on characteristics of an isomerous many-core hardware framework and characteristics of dynamic periodic changes during program operation, estimating optimal thread count in real time by acquiring state information of a system stage performance counter which reflects program performance changes through a predicating model; according to the optimal thread count, dynamically regulating the thread count of the application program. In order to reduce the system energy consumption while improving the program computational performances by reasonably utilizing computational resources of the isomerous many-core system, an effective solving method is provided.

Method for realizing multiple network device link aggregation

GPU-oriented data stream processing method

The invention discloses a graphics processing unit (GPU)-oriented data stream processing method. The method comprises the following steps of: operating a data-intensive computation module to be processed in a heterogeneous parallel system consisting of a central processing unit (CPU) and the GPU to acquire in-copy time Ti, computation time Tc and back-copy time To; computing the following proportional relation according to the three parts of running time: a is equal to Tc/Ti, and b is equal to Tc/To; and by combining the total amount of the initial data, determining the size of each data stream according to the summed size ratio among data streams when the initial data is divided into n streams, namely 1:a:a<2>:...:a:a/b:a/b2:...:a/b, and delivering the data streams to the GPU to perform in batches. By the method, the time for communication and computation is more efficiently superposed on the basis of avoiding system resource waste as much as possible, and the ...

Layered and clustered scheduling method, system, medium and equipment

The invention discloses a hierarchical clustering scheduling method and system, a medium and equipment, and the method comprises the steps: collecting various resource data of a cluster, and generating a node delay matrix; selecting proper delay hierarchical scheduling according to general tolerant calling response time of a scheduling task, searching whether the currently selected delay hierarchical exists or not, if not, clustering a graph formed by cluster nodes by using a complete sub-graph division method, hierarchically storing clustering information according to different delay standards after clustering is completed, and performing scheduling according to the clustering information. If so, reading clustering information of the current delay hierarchy; and then calculating the priority of each cluster according to a cluster priority formula, and selecting the cluster with high priority for scheduling. According to the method, the tasks are limited in the cluster to be scheduled, ...

Instant collection method of network resource accounting information based on event trigger mechanism

The invention discloses an instant collection method of network resource accounting information based on an event trigger mechanism. When a thread job Supervisor monitors the operation of finish execution in a trigger state, a network resource use record collector assembly in a network accounting system is activated, an operation inquire command supplied by an operation management system is called, wherein a return value of the command is a text containing resource amount information used in the operation execution process, the text is analyzed, accounting information needed to be collected is matched based on a special regular expression, and the collection of the resource amount used in the operation execution process can be finished; the accounting information is stored in a database and persistently stored; and finally, a state of the operation is converted to a completion state. If text analysis fails, an analysis task is executed again on the test returned by the operation inquire ...

A concurrent storage system construction method for convenient expansion of storage node quantity

This invention discloses a structuring method for a parallel storage system convenient for expanding numbers of stored nodes, which uses a data distribution mode of selecting in order to structure a storage system, when stored nodes are increased, it' s only necessary to move data from the original stored node to newly increased nodes, and it' s not necessary to move data among the original stored nodes to secure balance of the volume of stored nodes and consistency of the sequence selection mode.

CBEA (Cell Broadband Engine Architecture)-oriented transplanting method for isomorphic platform application

The invention discloses a CBEA (Cell Broadband Engine Architecture)-oriented transplanting method for isomorphic platform application. The method comprises the following steps of: finding a computing part of the entire application by adopting a subdivision-grade dynamic insertion subdivision technique on the basis of setting a guidance statement standard; and optimizing a computing core part of the application by adopting a 'Functional Offload' strategy through a source to source compiler of a compiling layer specific to an insertion guidance statement displayed by a computing core to bring the performance of an isomorphic multi-core framework into full play at a smaller transplanting overhead. By adopting a final runtime mechanism, the performance of the recompiled runtime of application specific to the isomorphic multi-core framework and the memory access characteristic of each computing part in the application can be acquired; and the application can be further optimized by a user according ...

Method for realizing two-way load equalizing mechanism in multiple machine servicer system

Fault-tolerant computer system data comparing method serving operating system core layer

The invention relates to a fault-tolerant computer system data comparing method serving an operating system core layer. The data comparing service is provided for a double-module redundancy process in a fault-tolerant computer system by starting a core daemon process in a Linux operating system and executing data comparator logics. An event list is added in a core to be used as a message passage; the redundancy process and a data comparator are operated in a production-consumption way; the redundancy process packages the data to be written into a message packet to be inserted into a message list; and the comparator takes the message packet from the message list, analyzes the message packet according to a definition format, compares the data to be written of the redundancy process and finally returns a result to the redundancy process. The invention is realized in the operating system core layer, has no customized hardware, simple realization and good generality and is suitable for a process-grade ...

Data direct connection method in AI computing power resource remote calling scene

A data direct connection method in an AI computing power resource remote calling scene comprises the following steps: an AI application in an AI computing power client initiates a data reading or preprocessing request, intercepts the data reading or preprocessing request and forwards the data reading or preprocessing request to a data service layer of an AI computing power server; the data service layer sends the data reading or preprocessing request to a storage service layer corresponding to the storage resource; the storage service layer accesses storage resources to perform data reading and preprocessing operation; an AI application in an AI computing power client initiates a request of copying data to computing power equipment, an AI calls a forwarding layer to intercept the request of copying the data to the computing power equipment and forward the request to a local data forwarding layer, and the data forwarding layer sends the request to a data service layer; the data service layer ...

Method for enhancing reliability of program of multi-replica contrast mechanism based on code segment

The invention discloses a method for enhancing the reliability of a program of a multi-replica contrast mechanism based on a code segment. In the method, an kernel thread is started as an initialization thread, the initialization thread obtains a process memory information, two replicas of a process code segment are stored in a physical memory, the process code segment is also divided into memory domains with the fixed size, the kernel threads with a corresponding amount are started as a consistency maintenance thread to maintain the consistency of a process code domain and a replica thereof in real time according to a division result, and program code segment errors caused by various software and hardware faults are discovered and recovered in time. The invention is simple to realize, is transparent to a protected process and can efficiently detect the program code segment errors, thereby preventing program crash and even system invalidation caused by the code segment errors and improving ...

Grid synchronization method for fault tolerant computer system based on socket

The invention discloses a network synchronization method based on a socket fault-tolerant computing system. A monitor is set on a socket interface layer of a Linux network architecture through modifying a Linux operating system so as to monitor the socket communication correlation kernel function. When the duplication redundancy schedule performs the network communication through the socket way, the communication operation is intercepted at the socket interface layer and the synchronization logic is triggered. The synchronization logic is implemented by reconstructing the socket series kernel functions, and the duplication schedule enters the synchronization point and completes the synchronizing process when implementing the reconstructed kernel functions. The implementation of the method is simple without special custom hardware. The reconstructed operating system is suitable for common hardware architectures with strong commonality. The synchronization process is initiatively completed ...

Parallel scheduling implementation method and system based on SHA-256 algorithm, medium and equipment

The invention discloses a parallel scheduling implementation method and system based on an SHA-256 algorithm, a medium and equipment, and designs parallel scheduling implementation of the SHA-256 algorithm for a SW 26010 processor. Through comprehensive performance optimization, including compilation option optimization, DMA transmission optimization, parallel scheduling optimization, instruction reduction optimization and assembly-level loop expansion and dual-emission technology optimization, efficient implementation of parallel scheduling of an SHA-256 algorithm is achieved. Compared with the implementation of an original SHA-256 algorithm with optimized memory access efficiency, the method has the advantages that the operation time is shortened from 45.01 s to 16.25 s for Hash calculation of 1MB messages, and the operation efficiency is improved by 2.77 times.

Single address traffic distributor of cluster network

The invention provides a single address flow sorter of group network. The sorter binds several physical terminals into a virtual physical terminal to extend the upstream flow, and allocates network flow to different processing joints according to the IP address of client terminal and upgrade the process ability. The data pack applying the IP is allocated with a different physical terminal, the process is realized with ASIC technology, the flow classification accordance is to work out Hash value to the client terminal IP address, because the data of the same user is allocated to the same physical terminal, thus it can provides continuous service to the joint connected to the physical joint.

Method for realizing dynamic layout of high-performance server based on group structure

Instant collection method of network resource accounting information based on event trigger mechanism

The invention discloses an instant collection method of network resource accounting information based on an event trigger mechanism. When a thread job Supervisor monitors the operation of finish execution in a trigger state, a network resource use record collector assembly in a network accounting system is activated, an operation inquire command supplied by an operation management system is called, wherein a return value of the command is a text containing resource amount information used in the operation execution process, the text is analyzed, accounting information needed to be collected is matched based on a special regular expression, and the collection of the resource amount used in the operation execution process can be finished; the accounting information is stored in a database and persistently stored; and finally, a state of the operation is converted to a completion state. If text analysis fails, an analysis task is executed again on the test returned by the operation inquire ...

Grid synchronization method for fault tolerant computer system based on socket

The invention discloses a network synchronization method based on a socket fault-tolerant computing system. A monitor is set on a socket interface layer of a Linux network architecture through modifying a Linux operating system so as to monitor the socket communication correlation kernel function. When the duplication redundancy schedule performs the network communication through the socket way, the communication operation is intercepted at the socket interface layer and the synchronization logic is triggered. The synchronization logic is implemented by reconstructing the socket series kernelfunctions, and the duplication schedule enters the synchronization point and completes the synchronizing process when implementing the reconstructed kernel functions. The implementation of the methodis simple without special custom hardware. The reconstructed operating system is suitable for common hardware architectures with strong commonality. The synchronization process is initiatively completed ...

Method for constructing high cost performance machine group network load equilization converter

This invention provides a method for constructing a high cost performance group network load balance converter utilizing the link aggregation function of the second exchange to bind multiple physical links in conformity with IEEE802.3ad standard to a logical link that multiple physical links have high usability and load balance at the same time and form a multiplayer high cost performance group network load balance converter.

Method for disposing micro-OS with self-adaption transmission model

The invention discloses a deployment method for a micro-operation system with a self-adaptive transmission model. With the method, the transmission model of the micro-operation system is adaptively constructed through comparing a node scale N in a cluster and the size of the node number M synchronously deployed by a TFTP server at most. When N is smaller than or equal to M, the micro-operation system is deployed by adopting a C/S transmission model; when M is smaller than N, and N is smaller than or equal to 2M, the nodes are deployed in batches by adopting the C/S transmission model, and the batching method overcomes the disadvantage that the deployment of the micro-operation system is failed caused by excessively heavy load of a TFTP server side; when N is larger than 2M, the C/S transmission model is automatically evolved into a multi-way tree transmission model, and by converting the nodes partially loaded with the micro-operation system into nodes which can provide TFTP service, the ...

Method for enhancing reliability of program of multi-replica contrast mechanism based on code segment

The invention discloses a method for enhancing the reliability of a program of a multi-replica contrast mechanism based on a code segment. In the method, an kernel thread is started as an initialization thread, the initialization thread obtains a process memory information, two replicas of a process code segment are stored in a physical memory, the process code segment is also divided into memorydomains with the fixed size, the kernel threads with a corresponding amount are started as a consistency maintenance thread to maintain the consistency of a process code domain and a replica thereof in real time according to a division result, and program code segment errors caused by various software and hardware faults are discovered and recovered in time. The invention is simple to realize, istransparent to a protected process and can efficiently detect the program code segment errors, thereby preventing program crash and even system invalidation caused by the code segment errors and improving ...