WO2020125427A1 - 针对递归树建模程序的资源关键树算法及系统 - Google Patents

针对递归树建模程序的资源关键树算法及系统 Download PDF

Info

Publication number
WO2020125427A1
WO2020125427A1 PCT/CN2019/123333 CN2019123333W WO2020125427A1 WO 2020125427 A1 WO2020125427 A1 WO 2020125427A1 CN 2019123333 W CN2019123333 W CN 2019123333W WO 2020125427 A1 WO2020125427 A1 WO 2020125427A1
Authority
WO
WIPO (PCT)
Prior art keywords
resource
target node
node
tree
resources
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2019/123333
Other languages
English (en)
French (fr)
Inventor
王洋
李敏
须成忠
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Institute of Advanced Technology of CAS
Original Assignee
Shenzhen Institute of Advanced Technology of CAS
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenzhen Institute of Advanced Technology of CAS filed Critical Shenzhen Institute of Advanced Technology of CAS
Publication of WO2020125427A1 publication Critical patent/WO2020125427A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/52—Program synchronisation; Mutual exclusion, e.g. by means of semaphores

Definitions

  • the invention belongs to the field of optoelectronic technology, and particularly relates to a resource key tree algorithm and system for a recursive tree modeling program.
  • the resource request program is often modeled as a recursive tree, which can express its inherent parallelism.
  • Each node in the recursive tree represents a computing task, and they request m types of resources.
  • Resources refer to various resources such as CPU, GPU, memory, hard disk, etc. required when the program is running, but improper allocation of resources will cause a deadlock of the program.
  • the execution of the recursive program is divided into two stages, the output tree and the input tree.
  • the output tree represents the execution of the tree node and requests for resource allocation. It must first complete the execution and resource allocation of its ancestor nodes, which is the stage of execution from top to bottom.
  • the input tree represents the completion of the release of computing tasks from bottom to top. The stage of returning in turn.
  • the purpose of the present invention is to provide a resource key tree algorithm for a recursive tree modeling program based on banker's algorithm, which aims to solve the problem that the existing resource request is not specifically designed according to the particularity of the recursive tree structure when it is modeled as a recursive tree
  • the technical problem of lock avoidance algorithm is to provide a resource key tree algorithm for a recursive tree modeling program based on banker's algorithm, which aims to solve the problem that the existing resource request is not specifically designed according to the particularity of the recursive tree structure when it is modeled as a recursive tree.
  • the present invention provides a resource key tree algorithm for a recursive tree modeling program, including the following steps:
  • the resource lower bound of the target node includes the minimum resource requirements of the target node and all its child nodes.
  • step S3 can allocate resources to the target node, then use the resource lower bound of the target node's parent node as the target node's relaxation resource.
  • the present invention also provides a resource key tree system for a recursive tree modeling program, including the following units:
  • the modeling unit is used to model the resource request program as a recursive tree model
  • a resource lower bound calculation unit connected to the modeling unit, is used to calculate the resource lower bound required by the target node to perform tasks and detect whether the resource request is safe;
  • a resource allocation unit connected to the resource lower bound calculation unit, is used to calculate the difference between the available resources and the resource lower bound, if the difference is positive, the target node is assigned resources; if the difference is negative, the resource is recovered To avoid deadlock;
  • the resource lower bound of the target node includes the minimum resource requirements of the target node and all its child nodes.
  • it also includes;
  • a relaxation resource calculation unit is connected to the resource allocation unit, and is configured to use the resource lower bound of the target node's parent node as the relaxation resource of the target node when the resource allocation unit allocates resources to the target node.
  • the resource key tree algorithm and system for the recursive tree modeling program of the present invention introduces the concept of a security detection algorithm into this algorithm.
  • the resource request of its branch structure is taken into consideration.
  • the lower bound of task execution resources for each node is raised, which improves the application scope and resource allocation performance of the resource requester in practical applications.
  • FIG. 1 is a basic algorithm structure diagram of a resource key tree algorithm for a recursive tree modeling program of the present invention
  • FIG. 2 is a complete algorithm structure diagram of the resource key tree algorithm for the recursive tree modeling program of the present invention
  • FIG. 3 is a functional unit block diagram of a resource key tree system for a recursive tree modeling program of the present invention
  • FIG. 4 is a schematic diagram of the recursive tree modeling program of the present invention.
  • 5 is a flowchart of the security detection algorithm of the present invention.
  • the airborne light-weight resource-critical tree algorithm for a recursive tree modeling program of the present invention includes the following steps:
  • the resource lower bound of the target node includes the minimum resource requirements of the target node and all its child nodes.
  • the security detection in the above resource key tree algorithm is to perform deadlock judgment or resource allocation according to the minimum resource that can prevent the node from deadlocking during deadlock detection.
  • step S3 can allocate resources to the target node, then use the resource lower bound of the target node's parent node as the target node's relaxation resource.
  • the security detection algorithm in step S2 includes the following steps:
  • u is the target node
  • v and v' are the left and right children of u in the input tree stage, respectively;
  • N(u) is the resource lower bound of the target node
  • ru is the resource required by the target node itself.
  • N'(u) is the relaxation resource of the target node
  • root is the root node
  • v is the parent node of the target node output tree stage
  • N(v) is the lower bound of the resource of the parent node of the target node output tree stage
  • rv is the target
  • the parent node itself needs resources in the node's output tree stage.
  • the present invention also provides a resource key tree system for a recursive tree modeling program, including the following units:
  • the modeling unit is used to model the resource request program as a recursive tree model
  • a resource lower bound calculation unit connected to the modeling unit, is used to calculate the resource lower bound required by the target node to perform tasks and detect whether the resource request is safe;
  • a resource allocation unit connected to the resource lower bound calculation unit, is used to calculate the difference between the available resources and the resource lower bound, if the difference is positive, the target node is assigned resources; if the difference is negative, the resource is recovered To avoid deadlock;
  • the resource lower bound of the target node includes the minimum resource requirements of the target node and all its child nodes.
  • it also includes;
  • a relaxation resource calculation unit is connected to the resource allocation unit, and is configured to use the resource lower bound of the target node's parent node as the relaxation resource of the target node when the resource allocation unit allocates resources to the target node.
  • the second embodiment is to encapsulate the algorithm of the first embodiment into a module or unit and add it to the resource scheduling system of the computer kernel.
  • the interface of the module allocates resources such as memory and hard disk to achieve the desired target effect.
  • the execution of the program includes two stages, a tree output tree stage, start each task calculation in a top-down manner (southbound), first allocate its required resources, and then in the possible In the case of starting its subtask thread, this process is recursively repeated by all its subtasks until the leaf task is completed; then another input tree stage is started, and the recursive process is gradually completed from the bottom to the top (northward) until the root task is completed. Therefore, the resource request graph of the program can be obtained by combining the output tree and input tree stages into a recursive tree. Given this model, the parent and child nodes of each node can be defined from the perspective of the output tree and the input tree.
  • the number of resources, the key tree for calculating resources, that is, the minimum resource N(u) that needs to be allocated for each node, is calculated as follows:
  • v and v' are the left and right children of u in the input tree stage, respectively, so that the minimum resource (resource lower bound) required for each node is calculated instead of the simple addition of the resources required by all child nodes.
  • node C (11,3) represents the resource ru requested by the node
  • ⁇ 36,10> represents the lower bound N(u) of the resource to be allocated by the node
  • node C left child N (u) is ⁇ 7,7>
  • the right child N(u) is ⁇ 25,2>
  • the ready queue Qr is assigned to the running variable Q and the sleep state node that may be released after the running state node is run is enqueued;
  • the node of queue Q cyclically dequeues, reclaims resources, determines whether the available resources can meet the resources required by the ready state node, and completes the security check;
  • N(u) is the resource lower bound of the completed subtree, and whenever the allocation of node u is checked in the real operation, the father node of its output tree has been allocated Resources, which can be used to estimate the relaxation resources of its child nodes (ie, the target node u).
  • N'(u) is the relaxation resource of the target node
  • root is the root node
  • v is the parent node of the target node output tree stage
  • N(v) is the lower bound of the resource of the parent node of the target node output tree stage
  • rv is the target
  • the parent node itself needs resources in the node's output tree stage.
  • the application program needs to perform resource scheduling when the system or the operating system needs to allocate resources for the recursive tree modeling program, call the module's interface to the memory, hard disk, etc. To allocate resources to achieve the desired target effect.
  • the concept of resource key algorithm is introduced. After fully considering the demand for execution resources of the branch structure of the tree, the execution resource of each node task is obtained Lower bound and avoid deadlock, allocate resources according to resource lower bound, greatly improve the efficiency of program execution.

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Debugging And Monitoring (AREA)

Abstract

一种适用于光电技术领域的针对递归树建模程序的资源关键树算法及系统，针对现有算法在资源请求可以建模为递归树的情况，没有根据递归树结构的特殊性设计专门的死锁的算法的问题，首先将资源请求程序建模为递归树模型；其次初始化安全检测算法，并采用该算法计算目标节点的执行任务所需的资源下界；再计算可用资源与所述资源下界的差值，若差值为正则为所述目标节点分配资源；若所述差值为负则回收资源，避免死锁；有效的提高了系统资源利用率和计算性能。

Description

针对递归树建模程序的资源关键树算法及系统 技术领域
本发明属于光电技术领域，尤其涉及一种针对递归树建模程序的资源关键树算法及系统。
背景技术
现有计算机系统资源分配时，资源请求程序常常建模为递归树，这样可以表示出其固有的并行性。递归树树中的每个结点都代表一个计算任务，他们请求m种类型的资源。资源是指程序运行的时候的需要cpu、gpu、内存、硬盘等各种资源，但是资源的不当分配会造成程序的死锁。递归程序的执行分为两个阶段，输出树和输入树。输出树代表树结点若要执行并请求资源分配，其必须先完成其祖先结点的执行和资源分配，是从上到下执行的阶段，输入树代表计算任务完成释放资源，从下到上的依次返回的阶段。当把每个递归树的结点即计算任务变为独立的线程或进程时，线程(进程)间可以并行执行来提高资源利用率和性能。
在资源充足的理想情况下，可以实现递归程序的并行最大化，但是在现实中，资源总是有限。递归树建模程序的两个阶段，其分叉和连接结构通常会给资源带来巨大压力，如现有资源无法满足节点请求的资源则引发死锁，甚至可能影响主要程序的运行和功能实现。现有技术存在不足。
发明内容
本发明的目的在于提供一种基于银行家算法的针对递归树建模程序的资源关键树算法，旨在解决现有资源请求建模为递归树时没有根据递归树结构的特殊性设计专门的死锁避免算法的技术问题。
本发明提供一种针对递归树建模程序的资源关键树算法，包括以下步骤：
S1.将资源请求程序建模为递归树模型；
S2.初始化安全检测算法，并采用该算法计算目标节点的执行任务所需的资源下界；
S3.计算可用资源与所述资源下界的差值，若差值为正则为所述目标节点分配资源；若所述差值为负则回收资源，避免死锁；
所述目标节点的资源下界包括目标节点及其所有子节点的最小资源需求。
优选的，还包括以下步骤；
S4.若步骤S3可以为所述目标节点分配资源，则利用所述目标节点的父节点的资源下界为所述目标节点的放松资源。
为了解决上述技术问题，本发明还提供一种针对递归树建模程序的资源关键树系统，包括以下单元：
建模单元，用于将资源请求程序建模为递归树模型；
资源下界计算单元，与所述建模单元连接，用于计算目标节点的执行任务所需的资源下界以及检测资源请求是否安全；
资源分配单元，与所述资源下界计算单元连接，用于计算可用资源与所述资源下界的差值，若差值为正则为所述目标节点分配资源；若所述差值为负则回收资源，避免死锁；
所述目标节点的资源下界包括目标节点及其所有子节点的最小资源需求。
优选的，还包括；
放松资源计算单元，与所述资源分配单元连接，用于当所述资源分配单元为所述目标节点分配资源时利用所述目标节点的父节点的资源下界为所述目标节点的放松资源。
本发明的针对递归树建模程序的资源关键树算法及系统，在此算法中引入了安全检测算法的概念，在考虑树的一个目标节点时将其分支结构的资源请求一并纳入考虑，得出每个结点任务执行资源下界，在实际应用中提升了资源请求程序的适用范围和资源分配性能。
附图说明
图1是本发明的针对递归树建模程序的资源关键树算法的基本算法结构图；
图2是本发明的针对递归树建模程序的资源关键树算法的完整算法结构图；
图3是本发明的针对递归树建模程序的资源关键树系统的功能单元框图；
图4是本发明的递归树建模程序示意图；
图5本发明的安全检测算法的流程图。
具体实施方式
为了使本发明的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。
实施例一：
如附图1-5所示，本发明的机载轻型一种针对递归树建模程序的资源关键树算法，包括以下步骤：
S1.将资源请求程序建模为递归树模型；
S2.初始化安全检测算法，并采用该算法计算目标节点的执行任务所需的资源下界；
S3.计算可用资源与所述资源下界的差值，若差值为正则为所述目标节点分配资源；若所述差值为负则回收资源，避免死锁；
所述目标节点的资源下界包括目标节点及其所有子节点的最小资源需求。
上述资源关键树算法中的安全检测就是在死锁检测的时候，按照能防止该结点死锁的最小资源来进行死锁判断或者资源分配。
优选的，还包括以下步骤；
S4.若步骤S3可以为所述目标节点分配资源，则利用所述目标节点的父节点的资源下界为所述目标节点的放松资源。
这样做的原因是如果在资源分配的时候只考虑该结点能安全运行所需的最小资源并按照最小资源分配，多个子孙结点再进行死锁检测的时候有可能产生更多不安全状态，所以按照比最小资源大一点的资源量分配资源会减少不安全状态的次数提高性能，这是平衡资源利用率和时间性能之间矛盾的一种方式。
优选的，所述步骤S2中的安全检测算法包括以下步骤：
D1.将就绪队列中的所述目标节点赋值给局部变量，
D2.遍历局部变量中的目标节点的所有子节点，
D3.将所有子节点调入运行队列；
D4.遍历所述运行队列的每个结点，当运行队列中的所有节点都被执行时，统计所需资源作为目标节点的最小资源需求；当可用资源不足以满足运行队列 中节点的运行需求时，返回不安全信号。
优选的，所述资源下界采用以下表达式计算：N(u)＝ru+MAX{N(v),N(v')}；
其中，u为所述目标节点，v和v'分别是u在输入树阶段的左孩子和右孩子；
N(u)为所述目标节点的资源下界，ru为所述目标节点自身所需资源。
优选的，若所述目标节点为根节点，其放松资源采用以下表达式计算：N'(u)＝N(root)；若所述目标节点不为根节点，其放松资源采用以下表达式计算：N'(u)＝N(v)-rv；
其中，N'(u)为目标节点的放松资源，root为根节点；v是目标节点输出树阶段的父节点，N(v)为目标节点输出树阶段的父节点的资源下界，rv为目标节点的输出树阶段的父节点自身所需资源。
实施例二：
如附图3所示，为了解决上述技术问题，本发明还提供一种针对递归树建模程序的资源关键树系统，包括以下单元：
建模单元，用于将资源请求程序建模为递归树模型；
资源下界计算单元，与所述建模单元连接，用于计算目标节点的执行任务所需的资源下界以及检测资源请求是否安全；
资源分配单元，与所述资源下界计算单元连接，用于计算可用资源与所述资源下界的差值，若差值为正则为所述目标节点分配资源；若所述差值为负则回收资源，避免死锁；
所述目标节点的资源下界包括目标节点及其所有子节点的最小资源需求。
优选的，还包括；
放松资源计算单元，与所述资源分配单元连接，用于当所述资源分配单元为所述目标节点分配资源时利用所述目标节点的父节点的资源下界为所述目 标节点的放松资源。
实施例二为将实施例一的算法封装成模块或单元，加入计算机内核的资源调度系统中，应用程序需要进行资源调度时通系统或者操作系统需要为递归树建模程序分配资源时，调用该模块的接口，对内存、硬盘等进行资源分配以达到理想的目标效果。
实施例三：
为了更好的说明本发明的技术方案，以下为具体实施时的详细步骤；
如附图4所示，程序的执行包括两个阶段，一个树输出树阶段，以自上而下的方式(南行)开始每个任务计算，首先分配其所需的资源，然后在可能的情况下启动其子任务线程，此过程由其所有子任务递归重复，直到叶子任务完成；然后启动另一个输入树阶段，从底部到顶部(北向)逐步完成递归过程，直到根任务完成。因此，通过将输出树和输入树阶段组合成递归树可以获得程序的资源请求图。给定这个模型，每个节点的父节点和子节点都可以从输出树和输入树角度定义，例如图1，对于结点C，从输出树阶段看A是C的父亲则有Pout(C)＝A，H和F是C的孩子则有Cout(C)＝H或者Cout(C)＝F，从输入树阶段看A是C的孩子则有Cin(C)＝A，同理有点Pin(C)＝H或者Pin(C)＝F，显然Pout(u)＝Cin(u)且Pin(u)＝Cout(u)。
关于执行递归树建模程序的资源的下限，每个结点u独立所需资源抽象为m维ru＝<r1,r2...rm>，m代表资源种类数，ri代表某种资源i需要的资源数，计算资源关键树，即每个结点需要分配的最少资源N(u)，计算方如下：
N(u)＝ru+MAX{N(v),N(v')}
其中，v和v'分别是u在输入树阶段的左孩子和右孩子，这样计算出每个结点所需分配的最小资源(资源下界)而不是所有子节点所需资源的简单相加。
如附图4所示，对于结点C，(11,3)表示该节点请求的资源ru，<36,10>表示该节点的需要分配的资源下界N(u),C结点左孩子N(u)为<7,7>，右孩子N(u)为<25,2>，所以N(u)＝<25+11,7+3>＝<36,10>，这是保证该结点顺利运行的最小应该分配的资源量。
如附图5所示，首先，开始并初始化安全检测算法；
其次，进入结点u的死锁检测算法并预分配资源；
再次，进入安全检测算法；
进一步的，就绪队列Qr赋值给运行变量Q并将运行状态结点运行后可能释放的睡眠状态的结点入队；
在一步的，队列Q的结点循环出队，回收资源，判断可用资源是否可以满足就绪状态节点所需资源，完成安全检测；
最后，若安全则确认分配资源，否则给该节点会造成死锁，回收资源。
关于放松资源：
虽然所提出的算法可以通过快速有效地解决递归树建模请求的死锁问题来成功实现我们的目标，但其实际应用仍然是一个巨大的挑战，因为：1)树可能随时间动态变化。2)在算法可以工作之前，必须知道与每个节点相关联的最大需求向量。
所以，即便如此在递归树中，我们认为仍然可以以相对简单的方式有效地实现算法。我们通过对于任何以目标节点u为根的子树，N(u)是完成子树的资源下界，并且真实运行中每当检查到节点u的分配时，其输出树的父亲结点已经分配过资源，可以利用它来估计其子节点(即目标节点u)的放松资源。
所以，如果我们将最大需求N(u)放宽到N'(u)，递归计算如下：
若所述目标节点为根节点，其放松资源采用以下表达式计算： N'(u)＝N(root)；若所述目标节点不为根节点，其放松资源采用以下表达式计算：N'(u)＝N(v)-rv；
其中，N'(u)为目标节点的放松资源，root为根节点；v是目标节点输出树阶段的父节点，N(v)为目标节点输出树阶段的父节点的资源下界，rv为目标节点的输出树阶段的父节点自身所需资源。
这样我们不仅能够包容死锁解决方案的请求信息的不足，而且能够摆脱树中每个目标节点u的N(u)计算，从而大大降低了安全检查的复杂性从O(mn log log n)到O(m log log n)，这放松资源算法得到非常有效的实现。
关于系统装置的实施部署：
将该算法封装成模块，加入计算机内核的资源调度系统中，应用程序需要进行资源调度时通系统或者操作系统需要为递归树建模程序分配资源时，调用该模块的接口，对内存、硬盘等进行资源分配以达到理想的目标效果。
在本发明实施例中，在计算某个结点需要分配的资源时，引入了资源关键算法的概念，在充分考虑树的分支结构对执行资源的需求后，得出每个结点任务执行资源下界并避免死锁，根据资源下界对资源进行分配，大大的提高了程序执行的效率。
以上所述仅为本发明的较佳实施例而已，并不用以限制本发明，凡在本发明的精神和原则之内所作的任何修改、等同替换和改进等，均应包含在本发明的保护范围之内。

Claims (7)

  1. 一种针对递归树建模程序的资源关键树算法，其特征在于，包括以下步骤：
    S1.将资源请求程序建模为递归树模型；
    S2.初始化安全检测算法，并采用该算法计算目标节点的执行任务所需的资源下界；
    S3.计算可用资源与所述资源下界的差值，若差值为正则为所述目标节点分配资源；若所述差值为负则回收资源，避免死锁；
    所述目标节点的资源下界包括目标节点及其所有子节点的最小资源需求。
  2. 如权利要求1所述的资源关键树算法，其特征在于，还包括以下步骤；
    S4.若步骤S3可以为所述目标节点分配资源，则利用所述目标节点的父节点的资源下界为所述目标节点的放松资源。
  3. 如权利要求2所述的资源关键树算法，其特征在于，所述步骤S2中的安全检测算法包括以下步骤：
    D1.将就绪队列中的所述目标节点赋值给局部变量，
    D2.遍历局部变量中的目标节点的所有子节点，
    D3.将所有子节点调入运行队列；
    D4.遍历所述运行队列的每个结点，当运行队列中的所有节点都被执行时，统计所需资源作为目标节点的最小资源需求；当可用资源不足以满足运行队列中节点的运行需求时，返回不安全信号。
  4. 如权利要求3所述的平台，其特征在于，所述资源下界采用以下表达式计算：N(u)＝ru+MAX{N(v),N(v')}；
    其中，u为所述目标节点，v和v'分别是u在输入树阶段的左孩子和右孩子；N(u)为所述目标节点的资源下界，ru为所述目标节点自身所需资源。
  5. 如权利要求4所述的资源关键树算法，其特征在于，若所述目标节点为根节点，其放松资源采用以下表达式计算：N'(u)＝N(root)；若所述目标节点不为根节点，其放松资源采用以下表达式计算：N'(u)＝N(v)-rv；
    其中，N'(u)为目标节点的放松资源，root为根节点；v是目标节点输出树阶段的父节点，N(v)为目标节点输出树阶段的父节点的资源下界，rv为目标节点的输出树阶段的父节点自身所需资源。
  6. 一种针对递归树建模程序的资源关键树系统，其特征在于，包括以下单元：
    建模单元，用于将资源请求程序建模为递归树模型；
    资源下界计算单元，与所述建模单元连接，用于计算目标节点的执行任务所需的资源下界以及检测资源请求是否安全；
    资源分配单元，与所述资源下界计算单元连接，用于计算可用资源与所述资源下界的差值，若差值为正则为所述目标节点分配资源；若所述差值为负则回收资源，避免死锁；
    所述目标节点的资源下界包括目标节点及其所有子节点的最小资源需求。
  7. 如权利要求6所述的资源关键树系统，其特征在于，还包括；
    放松资源计算单元，与所述资源分配单元连接，用于当所述资源分配单元为所述目标节点分配资源时利用所述目标节点的父节点的资源下界为所述目标节点的放松资源。
PCT/CN2019/123333 2018-12-18 2019-12-05 针对递归树建模程序的资源关键树算法及系统 Ceased WO2020125427A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811546982.9 2018-12-18
CN201811546982.9A CN109815021B (zh) 2018-12-18 2018-12-18 针对递归树建模程序的资源关键树方法及系统

Publications (1)

Publication Number Publication Date
WO2020125427A1 true WO2020125427A1 (zh) 2020-06-25

Family

ID=66602113

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/123333 Ceased WO2020125427A1 (zh) 2018-12-18 2019-12-05 针对递归树建模程序的资源关键树算法及系统

Country Status (2)

Country Link
CN (1) CN109815021B (zh)
WO (1) WO2020125427A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109815021B (zh) * 2018-12-18 2021-03-23 深圳先进技术研究院 针对递归树建模程序的资源关键树方法及系统
CN110347516B (zh) * 2019-06-27 2023-03-24 河北科技大学 一种面向细粒度读写锁的软件自动重构方法及装置
CN113360290B (zh) * 2020-03-04 2023-02-03 华为技术有限公司 死锁检测方法和装置
CN112486598A (zh) * 2020-10-28 2021-03-12 武汉中科通达高新技术股份有限公司 一种使用管道技术处理图片的方法、系统及电子装置

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7703102B1 (en) * 1999-08-23 2010-04-20 Oracle America, Inc. Approach for allocating resources to an apparatus based on preemptable resource requirements
CN105072049A (zh) * 2015-08-31 2015-11-18 北京理工大学 面向数据中心多层次弹性应用的资源分配方法及装置
CN106528287A (zh) * 2015-09-09 2017-03-22 阿里巴巴集团控股有限公司 计算机系统资源分配方法和装置
CN109815021A (zh) * 2018-12-18 2019-05-28 深圳先进技术研究院 针对递归树建模程序的资源关键树算法及系统

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6381742B2 (en) * 1998-06-19 2002-04-30 Microsoft Corporation Software package management
US8091088B2 (en) * 2005-02-22 2012-01-03 Microsoft Corporation Method and system for hierarchical resource management involving hard and soft resource limits
CN101394656B (zh) * 2007-09-21 2013-10-09 电信科学技术研究院 一种共享信道的资源分配方法及装置
CN103473137B (zh) * 2013-09-16 2017-04-12 东软集团股份有限公司 避免死锁的资源分配方法及系统
US9417918B2 (en) * 2013-11-20 2016-08-16 International Business Machines Corporation Computing session workload scheduling and management of parent-child tasks

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7703102B1 (en) * 1999-08-23 2010-04-20 Oracle America, Inc. Approach for allocating resources to an apparatus based on preemptable resource requirements
CN105072049A (zh) * 2015-08-31 2015-11-18 北京理工大学 面向数据中心多层次弹性应用的资源分配方法及装置
CN106528287A (zh) * 2015-09-09 2017-03-22 阿里巴巴集团控股有限公司 计算机系统资源分配方法和装置
CN109815021A (zh) * 2018-12-18 2019-05-28 深圳先进技术研究院 针对递归树建模程序的资源关键树算法及系统

Also Published As

Publication number Publication date
CN109815021A (zh) 2019-05-28
CN109815021B (zh) 2021-03-23

Similar Documents

Publication Publication Date Title
WO2020125427A1 (zh) 针对递归树建模程序的资源关键树算法及系统
Kulkarni et al. Survey on Hadoop and Introduction to YARN.
US9367359B2 (en) Optimized resource management for map/reduce computing
CN113205417B (zh) 一种面向联盟链的交易并行处理方法及系统
CN114741207A (zh) 一种基于多维度组合并行的gpu资源调度方法和系统
CN112612613B (zh) 用于在虚拟分布式异构环境下实现多gpu调度的方法和系统
CN107291550B (zh) 一种针对迭代应用的Spark平台资源动态分配方法及系统
CN104536937A (zh) 基于cpu-gpu异构集群的大数据一体机实现方法
CN111522640A (zh) 计算图的并行执行方法和设备
CN107220111A (zh) 一种基于任务窃取的任务调度方法及系统
CN112948066A (zh) 一种基于异构资源的Spark任务调度方法
US10102098B2 (en) Method and system for recommending application parameter setting and system specification setting in distributed computation
CN120196421A (zh) 一种gpu资源虚拟化算力调度的方法及装置
Zhao et al. Minimizing stack memory for partitioned mixed-criticality scheduling on multiprocessor platforms
CN113051049B (zh) 任务调度系统、方法、电子设备及可读存储介质
CN118051331A (zh) 资源分配方法及系统
CN107870813A (zh) 一种分布式算法处理数据的方法及装置
CN105607955A (zh) 一种计算任务分配的方法及装置
Das et al. Data races and the discrete resource-time tradeoff problem with resource reuse over paths
CN113934525A (zh) 一种基于正负反馈负载调度算法的Hadoop集群任务调度方法
CN110245022B (zh) 海量数据下并行Skyline处理方法及系统
CN119783340A (zh) 基于负载均衡的并行仿真方法、介质及设备
CN112463397B (zh) lock-free的分布式死锁避免方法及装置、计算机设备及可读存储介质
CN106775925A (zh) 一种虚拟机cpu的限额处理方法和装置
Wang et al. On optimal budget-driven scheduling algorithms for MapReduce jobs in the hetereogeneous cloud

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19898914

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19898914

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 11.11.2021)

122 Ep: pct application non-entry in european phase

Ref document number: 19898914

Country of ref document: EP

Kind code of ref document: A1