CN103365717B - Memory pool access method, Apparatus and system - Google Patents
Memory pool access method, Apparatus and system Download PDFInfo
- Publication number
- CN103365717B CN103365717B CN201310257057.5A CN201310257057A CN103365717B CN 103365717 B CN103365717 B CN 103365717B CN 201310257057 A CN201310257057 A CN 201310257057A CN 103365717 B CN103365717 B CN 103365717B
- Authority
- CN
- China
- Prior art keywords
- node
- internal memory
- memory
- access
- operating system
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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]
- G06F9/5083—Techniques for rebalancing the load in a distributed system
- G06F9/5088—Techniques for rebalancing the load in a distributed system involving task migration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/50—Indexing scheme relating to G06F9/50
- G06F2209/502—Proximity
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/50—Indexing scheme relating to G06F9/50
- G06F2209/508—Monitor
Abstract
The embodiment of the present invention provides a kind of memory pool access method, Apparatus and system, wherein memory pool access method includes that Node Controller receives the monitoring information that operating system sends, monitoring information carries the information of monitored internal memory in first node belonging to Node Controller, monitored internal memory is the memory source that target process takies on first node, and target process is the process that operation and the internal memory to the access node outside first node in server system conduct interviews on the central processor CPU of first node；If monitoring frequency that the internal memory of access node conducts interviews by the target process taking monitored internal memory more than or equal to threshold value, then the information of access node is sent to operating system, with the information according to access node, target process is migrated to access node；Remote memory is accessed and is converted to local internal storage access or neighbouring internal storage access such that it is able to reduce target process and access the time of internal memory, be effectively improved the performance of server system.
Description
Technical field
The present embodiments relate to computer technology, particularly relate to a kind of memory pool access method, device and be
System.
Background technology
Along with the development of computer technology, server system can be made up of one or more server,
Every station server, as a node, forms Non Uniform Memory Access and accesses (Non-Uniform Memory
Access, NUMA) framework.
In the server system of NUMA architecture, each node can include one or more central authorities
Processing unit (Central Processing Unit, CPU), each CPU can be configured in advance
Certain memory source.The process of the upper operation of CPU can use following three kinds of methods to server system
In memory source conduct interviews, be respectively local (local) internal storage access, neighbouring (buddy)
Internal storage access and far-end (remote) internal storage access.
The method that the memory source of this CPU self is conducted interviews by the process of the upper operation of CPU, i.e. local
Internal storage access；The memory source of other CPU in node belonging to this CPU is entered by the process of the upper operation of CPU
The method that row accesses, i.e. adjacent to internal storage access；The process of the upper operation of CPU to node belonging to this CPU with
The method that memory source outer, CPU in other nodes conducts interviews, i.e. remote memory accesses.
In these three memory pool access method, the process of the upper operation of CPU carries out remote memory and accesses required
Time longer, may be therefore, existing for carrying out 3-20 times of local internal storage access required time
The process run on CPU in server system is at the way access internal memory using remote memory to access
During resource, the overall performance of system can be caused to reduce.
Summary of the invention
The embodiment of the present invention provides a kind of memory pool access method, Apparatus and system, is used for improving server system
The performance that system is overall.
First aspect of the embodiment of the present invention is to provide a kind of memory pool access method, including:
Node Controller receives the monitoring information that operating system sends, and carries described in described monitoring information
The information of monitored internal memory in first node belonging to Node Controller, described monitored internal memory is target process
The memory source taken on described first node, described operating system is by including described first segment
On the point each node in the server system of interior at least two node composition, described target process is
The central processor CPU of described first node runs and to described in described server system first
The process that the internal memory of the access node outside node conducts interviews；
If described Node Controller monitors takies the described target process of described monitored internal memory to described quilt
The frequency that the internal memory of access node conducts interviews is more than or equal to threshold value, then by the letter of described access node
Breath is sent to described operating system so that described operating system according to the information of described access node by institute
State target process and migrate to described access node.
The memory pool access method provided in conjunction with first aspect, in the implementation that the first is possible, institute
Stating threshold value is in the process run on whole CPU in described first node in Preset Time, saves other
The number of times that the internal memory of point conducts interviews and the ratio of CPU sum in described first node.
In conjunction with first aspect or implementation that the first is possible, in the implementation that the second is possible,
Described Node Controller is Node Controller NC chip；
Correspondingly, the monitoring information that described Node Controller reception operating system sends includes:
The described Node Controller NC chip mainboard management control unit by described first node, receives institute
State the described monitoring information that operating system sends；
Correspondingly, if described Node Controller monitors the described target process taking described monitored internal memory
The frequency conducting interviews the internal memory of described access node is more than or equal to threshold value, then by described accessed
The information of node is sent to described operating system and includes:
Described Node Controller NC chip, if monitoring the described target process taking described monitored internal memory
The frequency conducting interviews the internal memory of described access node is more than or equal to described threshold value, then by described quilt
Access the information of accessed internal memory in node, be sent to described operation by described mainboard management control unit
System.
Second aspect of the embodiment of the present invention is to provide a kind of Node Controller, including:
Receive unit, for receiving the monitoring information that operating system sends, described monitoring information carries
The information of monitored internal memory in first node belonging to described Node Controller, described monitored internal memory is target
The memory source that process takies on described first node, described operating system is by including described
On one node each node in the server system of interior at least two node composition, described target is entered
Journey is to run and to described in described server system on the central processor CPU of described first node
The process that the internal memory of the access node outside first node conducts interviews；
Monitoring unit, for monitoring the described target process of described monitored internal memory to described accessed
The frequency that the internal memory of node conducts interviews is more than or equal to threshold value, then the information of described access node sent out
Give described operating system so that described operating system according to the information of described access node by described mesh
Mark process migration is to described access node.
The Node Controller provided in conjunction with second aspect, in the implementation that the first is possible, described
Threshold value is in the process run on whole CPU in described first node in Preset Time, to other nodes
The number of times that conducts interviews of internal memory and the ratio of CPU sum in described first node.
In conjunction with second aspect or implementation that the first is possible, in the implementation that the second is possible,
Described reception unit specifically for:
By the mainboard management control unit of described first node, receive the described of described operating system transmission
Monitoring information；
Described monitoring unit specifically for:
Monitoring the described target process taking described monitored internal memory, the internal memory of access node is carried out
When the frequency accessed is more than or equal to described threshold value, by the information of internal memory accessed in described access node,
It is sent to described operating system by described mainboard management control unit.
3rd aspect of the embodiment of the present invention is to provide a kind of server system, includes including at least two
The node of Node Controller；Described server system runs and has operating system；
If described operating system is judged to there is target process, then run the of place to described target process
One node sends monitoring information, carries described the belonging to described Node Controller in described monitoring information
The information of monitored internal memory in one node, described monitored internal memory is that described target process is at described first segment
The memory source taken on point, described target process is on the central processor CPU of described first node
Run and the internal memory of the access node outside first node described in described server system is visited
The process asked；
What what described operating system received that described first node sends accessed by described target process is accessed
After the information of node, described target process is migrated to described access node.
The server system provided in conjunction with the 3rd aspect, in the implementation that the first is possible, described
If operating system judges accessed CPU belonging to internal memory in described access node, its memory source has
There is the ability running described target process, then described target process is migrated to belonging to described accessed internal memory
CPU；If judging the CPU belonging to described accessed internal memory, its memory source does not have the described mesh of operation
The memory source of other CPU in the ability of mark process, and described access node has the described mesh of operation
The ability of mark process, then migrate to other CPU of described access node by described target process.
The memory pool access method of embodiment of the present invention offer, Apparatus and system, Node Controller is receiving
After the monitoring information that operating system sends, if monitoring the process taking monitored internal memory to accessed joint
The internal memory of point carries out the frequency of remote access and is more than or equal to threshold value, then the information of access node sent
To operating system, so that target process is migrated to described access node by operating system；By remote memory
Access is converted to local internal storage access or neighbouring internal storage access such that it is able to reduces target process and accesses internal memory
Time, be effectively improved the performance of server system.
Accompanying drawing explanation
The flow chart of the memory pool access method that Fig. 1 provides for the embodiment of the present invention；
The flow chart of another memory pool access method that Fig. 2 provides for the embodiment of the present invention；
The structural representation of the Node Controller that Fig. 3 provides for the embodiment of the present invention；
The structural representation of another Node Controller that Fig. 4 provides for the embodiment of the present invention；
The structural representation of the server system that Fig. 5 provides for the embodiment of the present invention.
Detailed description of the invention
Various embodiments of the present invention in order to improve process use remote memory access mode time, server system
The performance of system, have employed the memory pool access method migrating the process meeting certain condition, by far-end
The mode of internal storage access is converted to local internal storage access or the mode of neighbouring internal storage access.
Various embodiments of the present invention can be applied in the multi-server system being made up of at least two node, respectively
Node includes the Node Controller described in various embodiments of the present invention.Each node can be a station server,
Can include that one or more CPU, each CPU can be assigned part accordingly on every station server
Memory source.Operating system (Operating System, OS) may operate in this server system.
The flow chart of the memory pool access method that Fig. 1 provides for the embodiment of the present invention, as it is shown in figure 1, the party
Method includes:
101, Node Controller receives the monitoring information that operating system sends.
Wherein, described monitoring information carry in first node belonging to described Node Controller monitored interior
The information deposited, described monitored internal memory is the memory source that target process takies on described first node,
Described operating system is at the server being made up of at least two node including described first node
On each node in system, described target process is on the central processor CPU of described first node
Run and the internal memory of the access node outside first node described in described server system is visited
The process asked.
Concrete, operating system is on each node of server system, on the CPU of each node
All may run one or more process, these processes can use local internal storage access, neighbouring internal memory
Access or remote memory accesses and carries out internal storage access.
Various embodiments of the present invention are mainly for carrying out the process of remote memory access, and therefore the present invention respectively implements
" target process " described in example i.e. refers to carry out the process of remote memory access.It is to say, target
Memory sources beyond node belonging to its CPU run on, in other nodes are visited by process
Ask." first node " described in various embodiments of the present invention is i.e. as the Node Controller institute of executive agent
The node belonged to, is so termed to make a distinction with other nodes, to avoid confusion.
The process run each node and each node can be managed by operating system from the angle of the overall situation
Reason, the ruuning situation of the process during therefore operating system can know each node.Such as, which node has
A little processes are being run, and which CPU is these processes operate on, and take is the interior of which CPU
Deposit.Operating system can know each process for internal memory the in the case of of conducting interviews, including process access be
Any partial memory resource on which CPU of which node.Therefore, remote memory access is carried out when process
Time, operating system can know this situation.
Operating system is in knowing the process that server system is run, when there is target process, to operation
Node Controller on the first node of this target process sends monitoring information.Wherein, monitoring information is wrapped
Include the information of the memory source that this target process takies on this first node, will in various embodiments of the present invention
This partial memory resource is referred to as " monitored internal memory ".
102, Node Controller is monitoring the described target process taking described monitored internal memory to accessed
When the frequency that the internal memory of node conducts interviews is more than or equal to threshold value, the information of described access node is sent out
Give described operating system.Performing step 102 is to make described operating system according to described accessed joint
The information of point, migrates to described access node by described target process.
Concrete, Node Controller is after receiving monitoring information, to quilt indicated in monitoring information
Monitoring internal memory is monitored.First node can know that the process run on it accesses the situation of internal memory.
Process can not be directly monitored by Node Controller, will be owing to operating in process on first node
This locality takies the partial memory on run on CPU, and therefore Node Controller can be to this partial memory
It is monitored.
Further, when the internal memory on other nodes is conducted interviews by the process on this first node, node control
Device processed also is able to know the information being accessed for internal memory on other nodes, by target in various embodiments of the present invention
Node belonging to the internal memory of process remote access, is referred to as " access node "；Will be located on other nodes
, the memory source accessed by target process, be referred to as " accessed internal memory ".
The Node Controller process to being run in monitored internal memory, conducts interviews to the internal memory of other nodes
Situation be monitored.It is specifically as follows the frequency that the internal memory of other nodes is conducted interviews by monitoring objective process
Degree.
Node Controller can be preset measurement period and threshold value.Node Controller monitors each measurement period
In, the number of times that the internal memory of access node is conducted interviews by destination node, it is thus achieved that frequency value.Node control
If device judges that this frequency value more than or equal to this threshold value, then illustrates that this target process mainly uses far
The mode that end memory accesses.In order to improve the performance of server system, in this case, node control
The information of access node is sent to operating system by device processed.Wherein, permissible in the information of access node
Including, the information of the accessed internal memory accessed by target process in access node.But, due to operation
System can know which the memory source that target process carries out being accessed during remote access is, therefore at quilt
Access node information in carry accessed internal memory information it is not necessary to.
After operating system receives the information of this access node, target process can be moved to this quilt
Access node, so that after Qian Yiing, target process can use local IP access or the neighbouring side accessed
Formula, conducts interviews to the internal memory of access node such that it is able to be effectively improved systematic function.
Further, target process can not also directly be migrated by operating system, to accessed joint
After the residue situation of the memory source in point judges, when remaining memory source enough runs this mesh
During mark process, then this target process is migrated, preferably to improve systematic function.
Wherein, the method that process is migrated by operating system, can use similar with prior art
Implementation, this repeats no more.
The memory pool access method that the embodiment of the present invention provides, Node Controller is receiving operating system transmission
Monitoring information after, if monitoring the target process taking monitored internal memory to the internal memory of access node
The frequency carrying out remote access is more than or equal to threshold value, then the information of access node is sent to operation system
System, so that target process is migrated to described access node by operating system；Remote memory is accessed conversion
For local internal storage access or neighbouring internal storage access such that it is able to reduce target process and access the time of internal memory,
It is effectively improved the performance of server system.
The flow chart of another memory pool access method that Fig. 2 provides for the embodiment of the present invention, as in figure 2 it is shown,
The method includes:
201, the Node Controller NC chip in first node passes through mainboard management control unit, receives behaviour
Make the monitoring information that system sends.
Concrete, on the basis of the most described implementation, between NC chip and operating system
Communication, be to be realized by mainboard management control unit set on first node.
202, the NC chip in first node enters monitoring the described target taking described monitored internal memory
When the frequency that the internal memory of described access node is conducted interviews by journey is more than or equal to described threshold value, by described
In access node, the information of accessed internal memory, is sent to described behaviour by described mainboard management control unit
Make system.
Concrete, may refer to the implementation described in step 102.
Further, in the server system of NUMA architecture, when the quantity of node is more, if adopted
Be attached node by the mode of direct interconnection, then the annexation between node will be limited by node
Interface quantity or interconnection communication protocol, therefore can use Node Controller (Node Controller,
NC) each node is interconnected by chip.It is provided with a NC chip on each node, divides between node
Tong Guo can not communicate to connect by respective NC chip, be belonging respectively between the CPU on two nodes to lead to
Cross NC chip and realize at a high speed interconnection, and the connection between node need not the interface that is limited to again on node
Quantity.NC chip has the functions such as caching (Cache) consistency check and message forwarding.
Operating system can be by mainboard management control unit (the Baseboard Management on node
Controller, BMC), the register on the NC chip on node is written and read operation.BMC supports
IPMI (the Intelligent Platform Management of professional standard
Interface, IPMI) specification.This specification describes the management function being built on mainboard, including
Local and remote diagnosis, console support, configuration management, hardware management and failture evacuation etc..IPMI is
The free standard of one opening being applied to server management system design.IPMI provides multiple systems interface,
KBC mode (Keyboard Controller Style, KCS) be currently used most widely
IPMI system interface.Operating system can be communicated by KCS with BMC.BMC can be by integrated
Circuit bus (Inter-Integrated Circuit, IIC) accesses the register of NC chip, carries out
Read-write operation, thus realize the monitoring to NC chip status or control.
Monitored internal memory is monitored by the NC chip on first node, and is transported on monitored internal memory
The situation that the internal memory of other nodes is conducted interviews by the target process of row, is monitored.
Further, described threshold value is to run on whole CPU in described first node in Preset Time
In process, number of times that the internal memory of other nodes is conducted interviews and the ratio of CPU sum in described first node
Value.
Concrete, threshold value can be set as required, and a kind of typical set-up mode is as follows.
First node is preset a time span, with NC chip on statistics Preset Time this first node interior
The process run on whole CPU of interconnection, uses the number of times that remote memory accesses, and it is right that this number of times had both included
The access of the shared drive in other nodes, also includes the access to the unshared internal memory in other nodes.
By the ratio of this number of times with the quantity of the CPU on this first node, as this threshold value.It is to say,
If the NC chip of first node monitors the target process run on monitored internal memory, carry out remote memory
The number of times accessed, more than or equal to carrying out the mean value of remote memory access on single cpu, then explanation should
Target process carries out remote memory and accesses more frequent, and this target process needs to be migrated.
Illustrate, such as the quantity of the CPU in node is 8, by this node in Preset Time
It is 1000 times that NC chip carries out the number of request of remote memory access, on i.e. 8 CPU run all enter
Journey has carried out 1000 remote memories altogether and has accessed.Thus can calculate, each CPU is right in Preset Time
The mean value of remote access is 125 times.
Thus, set the threshold to 125.When target process carries out remote memory access in measurement period
Number of times more than or equal to 125, then illustrate that this target process is more frequent to the access of remote memory, need
To be migrated.
The structural representation of the Node Controller that Fig. 3 provides for the embodiment of the present invention, as it is shown on figure 3, should
Node Controller includes:
Receive unit 11, for receiving the monitoring information that operating system sends, described monitoring information is carried
Having the information of monitored internal memory in first node belonging to described Node Controller, described monitored internal memory is mesh
The memory source that mark process takies on described first node, described operating system is described by including
On the first node each node in the server system of interior at least two node composition, described target
Process is to run and to institute in described server system on the central processor CPU of described first node
State the process that the internal memory of the access node outside first node conducts interviews；
Monitoring unit 12, for monitoring the described target process of described monitored internal memory to described accessed
The frequency that the internal memory of node conducts interviews is more than or equal to threshold value, then the information of described access node sent out
Give described operating system so that described operating system according to the information of described access node by described mesh
Mark process migration is to described access node.
Further, for entering with the described target taking described monitored internal memory in described monitoring unit 12
The described threshold value that the frequency that the internal memory of described access node is conducted interviews by journey compares, described threshold value
It is in the process run on whole CPU in described first node in Preset Time, in other nodes
Deposit the number of times and the ratio of CPU sum in described first node conducted interviews.
Further, described reception unit 11 specifically for:
By the mainboard management control unit of described first node, receive the described of described operating system transmission
Monitoring information；
Described monitoring unit 12 specifically for:
Monitoring the described target process taking described monitored internal memory, the internal memory of access node is carried out
When the frequency accessed is more than or equal to described threshold value, by the information of internal memory accessed in described access node,
It is sent to described operating system by described mainboard management control unit.
The structural representation of another Node Controller that Fig. 4 provides for the embodiment of the present invention, as shown in Figure 4,
This Node Controller includes:
Processor 21, memory 22, bus 23 and communication interface 24.Processor 21, memory 22
And connect by bus 23 and complete mutual communication between communication interface 24.
Processor 21 may be monokaryon or multinuclear CPU (Central Processing Unit,
CPU), or it is specific integrated circuit (Application Specific Integrated Circuit, ASIC),
Or for being configured to implement one or more integrated circuits of the embodiment of the present invention.
Memory 22 can be high-speed RAM memory, it is also possible to for nonvolatile memory
(non-volatile memory), for example, at least one magnetic disc store.
Memory 22 is used for program of depositing 221.Concrete, program 221 can include program code,
Described program code includes computer-managed instruction.
Communication interface 24, for receiving the monitoring information that operating system sends, carries in described monitoring information
The information of monitored internal memory in first node belonging to described Node Controller, described monitored internal memory is target
The memory source that process takies on described first node, described operating system is by including described
On one node each node in the server system of interior at least two node composition, described target is entered
Journey is to run and to described in described server system on the central processor CPU of described first node
The process that the internal memory of the access node outside first node conducts interviews.
Processor 21 runs program 221, to perform:
At the described target process monitoring described monitored internal memory, the internal memory of described access node is carried out
The information of described access node more than or equal to threshold value, is then sent to described operation system by the frequency accessed
System, so that described target process is migrated to institute according to the information of described access node by described operating system
State access node.
Concrete, that Node Controller that various embodiments of the present invention provide carries out internal storage access method is permissible
Seeing the operating procedure described in the embodiment of the method for above-mentioned correspondence, here is omitted.
The structural representation of the server system that Fig. 5 provides for the embodiment of the present invention, as it is shown in figure 5, should
Server system includes at least two node, respectively includes Node Controller as described in fig 3
Node 1 and node 2.
Described server system runs and has operating system；If described operating system judges that there is target enters
Journey, then send monitoring information, described monitoring to the first node controller that described target process is run on
Information carries monitored internal memory in node belonging to first node controller belonging to described Node Controller
Information, described monitored internal memory is the memory source that described target process takies on described first node,
Described target process is to run and to described server on the central processor CPU of described first node
The process that the internal memory of access node outside first node described in system conducts interviews；
What what described operating system received that described first node sends accessed by described target process is accessed
After the information of node, described target process is migrated to described access node.
Concrete, may refer to step 101-102, or the implementation described in 201-202.
Further, server system can be in the following way in actual applications, it is achieved to internal memory
Access.
Operating system operationally, if it is judged that there is process creation shared drive, will this be shared interior
Deposit as monitored internal memory, i.e. start shared drive monitoring module.Shared drive monitoring module, is used for sentencing
Whether this process disconnected conducts interviews to memory sources beyond the node affiliated at it, other nodes；If
It is, then os starting OS information reporting module.OS information reporting module, for monitoring information
It is sent to BMC.
Wherein, monitoring information includes No. CPU, node number and the information of monitored internal memory.No. CPU is CPU
Unique mark in whole server system；Node number is unique in whole server system of node
Mark；The information NC chip of monitored internal memory needs the description information of the region of memory of monitoring, if usually
The starting and ending address of dry group memory sections physical address.Such as, 1KB memory headroom, in physics
Depositing addressing and may have 2 sections, one section is 0x0-0x200, and another section is 0x400-0x600, then describe letter
Breath should be 2 groups of i.e. 0x0-0x200 and 0x400-0x600 of description.
Operating system may select KCS or monolithic transmission means (One-Block with the communication port of BMC
Transfer, BT)；In monitoring flow process, OS information reporting module sends information, and BMC to BMC
To operating system feedback information and migration instruction etc., all sent by KCS or BT mode.
After BMC receives monitoring information, utilize its modular converter that this monitoring information is converted to NC chip
The value of middle particular register, and by register module for reading and writing by the monitoring mesh of these information write NC chip
Scalar register file, is monitored by NC chip.
NC chip can be according to these register informations, by the monitoring module of its network subsystem, to being supervised
The frequency that internal memory on other nodes is conducted interviews by the process run on control internal memory, is monitored.When this
When frequency exceedes threshold value, start feedback module alarm, and interior with monitored by No. CPU of monitored CPU
The information deposited is sent to BMC, BMC be transmitted to operating system.
Wherein, need to increase threshold register in NC chip, be used for depositing threshold value；Increase monitoring to deposit
Device group, for depositing monitored No. CPU and node number, the Yi Jili corresponding with OS information reporting module
The memory sections information of monitored internal memory is deposited, it is also possible to include that target memory is visited with memory sections register
Ask counter, carry out the number of times of remote memory access for the process of adding up；Increase alarm register, be used for
Indicate whether that needs migrate；Increase feedback register, for being supplied to what BMC needed to carry out migrating
No. CPU, node number and the memory sections information that process is corresponding, the content information of its instruction is deposited with monitoring
Device content is corresponding.
It addition, the threshold value deposited in threshold register can be configured by operating system by BMC.
Target memory access counter can be compared by NC chip termly with the numerical value in threshold register,
To judge that process carries out the number of times of remote memory access and whether exceedes threshold value.This regularly time length permissible
It is configured as required, such as 60 seconds or 120 seconds etc., but optional numerical value is not limited to that.
BMC is read out by the numerical value in the timing register to NC chip, if judging, alarm is deposited
Device is set, then read the numerical value in feedback register, and turned by the numerical value of register by modular converter
Change No. CPU and memory sections information of correspondence into, and combine the instruction needing to migrate, be sent to operation system
System.BMC reads the time interval of alarm register and can be set as required, such as 60 seconds or 120
Second etc., but optional numerical value is not limited to that.
Operating system is after receiving migration instruction, according to No. CPU received and memory sections information,
Judge to need which process is migrated, and realized the migration of process by process migration module.
Further, if described operating system is judged to be accessed belonging to internal memory in described access node
CPU, its memory source has the ability running described target process, then described target process is migrated to institute
State the CPU belonging to accessed internal memory；If judging the CPU belonging to described accessed internal memory, its memory source
Not there is the ability running described target process, and the internal memory money of other CPU in described access node
Source has the ability running described target process, then described target process is migrated to described access node
Other CPU.
Concrete, in actual applications, operating system is after receiving migration instruction, it is also possible to enter one
Step ground judges destination node, i.e. access node, if having enough memory sources and can distribute to this mesh
Mark process.
If the free memory resource being accessed in internal memory in destination node is sufficiently used for distributing to this target and enters
Journey, i.e. has the ability running described target process, then target process can be migrated to this mesh by operating system
Node this partial memory in.
If the free memory inadequate resource in accessed internal memory is enough in and distributes to this target and enter in destination node
Journey, does not the most have the ability running described target process, then can not migrate.Can also continue
This partial memory is detected by continuous timing, and when free memory resource is enough, operating system can be by target
Process migration is in this partial memory of this destination node.
Operating system can also judge the free memory resource in destination node in the internal memory of other CPU, is
No being sufficiently used for distributes to this target process, if it is, target process can temporarily migrate to this portion
Divide internal memory, and continue regularly accessed internal memory to be detected, when the free memory in accessed internal memory provides
When source is sufficiently used for distributing to this target process, target process is migrated in this accessed internal memory.
The memory pool access method of various embodiments of the present invention offer, Apparatus and system, by process migration, can
The CPU of move target node, memory source to be fully utilized, improve the utilization rate of resource；
Simultaneously because process migration, if the process in node is moved out, and there is CPU that process moved out without other
Need the process run, then can reduce the energy consumption of this node, it is achieved energy-conservation.
One of ordinary skill in the art will appreciate that: realize all or part of step of above-mentioned each method embodiment
Suddenly can be completed by the hardware that programmed instruction is relevant.Aforesaid program can be stored in a computer can
Read in storage medium.This program upon execution, performs to include the step of above-mentioned each method embodiment；And
Aforesaid storage medium includes: ROM, RAM, magnetic disc or CD etc. are various can store program code
Medium.
Last it is noted that various embodiments above is only in order to illustrate technical scheme, rather than right
It limits；Although the present invention being described in detail with reference to foregoing embodiments, this area common
Skilled artisans appreciate that the technical scheme described in foregoing embodiments still can be modified by it,
Or the most some or all of technical characteristic is carried out equivalent；And these amendments or replacement, and
The essence not making appropriate technical solution departs from the scope of various embodiments of the present invention technical scheme.
Claims (8)
1. a memory pool access method, it is characterised in that including:
Node Controller receives the monitoring information that operating system sends, and carries described in described monitoring information
The information of monitored internal memory in first node belonging to Node Controller, described monitored internal memory is target process
The memory source taken on described first node, described operating system is by including described first segment
On the point each node in the server system of interior at least two node composition, described target process is
The central processor CPU of described first node runs and to described in described server system first
The process that the internal memory of the access node outside node conducts interviews；
If described Node Controller monitors takies the described target process of described monitored internal memory to described quilt
The frequency that the internal memory of access node conducts interviews is more than or equal to threshold value, then by the letter of described access node
Breath is sent to described operating system so that described operating system according to the information of described access node by institute
State target process and migrate to described access node.
Memory pool access method the most according to claim 1, it is characterised in that described threshold value is in advance
If in the process run on whole CPU in described first node in the time, the internal memory of other nodes being carried out
The number of times accessed and the ratio of CPU sum in described first node.
Memory pool access method the most according to claim 1 and 2, it is characterised in that described node control
Device processed is Node Controller NC chip；
Correspondingly, the monitoring information that described Node Controller reception operating system sends includes:
The described Node Controller NC chip mainboard management control unit by described first node, receives institute
State the described monitoring information that operating system sends；
Correspondingly, if described Node Controller monitors the described target process taking described monitored internal memory
The frequency conducting interviews the internal memory of described access node is more than or equal to threshold value, then by described accessed
The information of node is sent to described operating system and includes:
Described Node Controller NC chip, if monitoring the described target process taking described monitored internal memory
The frequency conducting interviews the internal memory of described access node is more than or equal to described threshold value, then by described quilt
Access the information of accessed internal memory in node, be sent to described operation by described mainboard management control unit
System.
4. a Node Controller, it is characterised in that including:
Receive unit, for receiving the monitoring information that operating system sends, described monitoring information carries
The information of monitored internal memory in first node belonging to described Node Controller, described monitored internal memory is target
The memory source that process takies on described first node, described operating system is by including described
On one node each node in the server system of interior at least two node composition, described target is entered
Journey is to run and to described in described server system on the central processor CPU of described first node
The process that the internal memory of the access node outside first node conducts interviews；
Monitoring unit, for monitoring the described target process of described monitored internal memory to described accessed
The frequency that the internal memory of node conducts interviews is more than or equal to threshold value, then the information of described access node sent out
Give described operating system so that described operating system according to the information of described access node by described mesh
Mark process migration is to described access node.
Node Controller the most according to claim 4, it is characterised in that described threshold value is to preset
In the process run on whole CPU in described first node in time, the internal memory of other nodes is visited
The number of times asked and the ratio of CPU sum in described first node.
6. according to the Node Controller described in claim 4 or 5, it is characterised in that described reception unit
Specifically for:
By the mainboard management control unit of described first node, receive the described of described operating system transmission
Monitoring information；
Described monitoring unit specifically for:
Monitoring the described target process taking described monitored internal memory, the internal memory of access node is carried out
When the frequency accessed is more than or equal to described threshold value, by the information of internal memory accessed in described access node,
It is sent to described operating system by described mainboard management control unit.
7. a server system, it is characterised in that include that at least two includes as in claim 4-6
The node of arbitrary described Node Controller；Described server system runs and has operating system；
If described operating system is judged to there is target process, then run the of place to described target process
One node sends monitoring information, carries described the belonging to described Node Controller in described monitoring information
The information of monitored internal memory in one node, described monitored internal memory is that described target process is at described first segment
The memory source taken on point, described target process is on the central processor CPU of described first node
Run and the internal memory of the access node outside first node described in described server system is visited
The process asked；
What what described operating system received that described first node sends accessed by described target process is accessed
After the information of node, described target process is migrated to described access node.
Server system the most according to claim 7, it is characterised in that if described operating system is sentenced
Breaking and accessed CPU belonging to internal memory in described access node, its memory source has the described mesh of operation
The ability of mark process, then migrate to the CPU belonging to described accessed internal memory by described target process；If sentencing
Breaking and the CPU belonging to described accessed internal memory, its memory source does not have the energy running described target process
The memory source of other CPU in power, and described access node has the energy running described target process
Power, then migrate to other CPU of described access node by described target process.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310257057.5A CN103365717B (en) | 2013-06-25 | 2013-06-25 | Memory pool access method, Apparatus and system |
| PCT/CN2014/071252 WO2014206078A1 (en) | 2013-06-25 | 2014-01-23 | Memory access method, device and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310257057.5A CN103365717B (en) | 2013-06-25 | 2013-06-25 | Memory pool access method, Apparatus and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103365717A CN103365717A (en) | 2013-10-23 |
| CN103365717B true CN103365717B (en) | 2016-08-31 |
Family
ID=49367142
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310257057.5A Active CN103365717B (en) | 2013-06-25 | 2013-06-25 | Memory pool access method, Apparatus and system |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN103365717B (en) |
| WO (1) | WO2014206078A1 (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103365717B (en) * | 2013-06-25 | 2016-08-31 | 华为技术有限公司 | Memory pool access method, Apparatus and system |
| CN104917784B (en) * | 2014-03-10 | 2018-06-05 | 华为技术有限公司 | A kind of data migration method, device and computer system |
| CN104035823B (en) * | 2014-06-17 | 2018-06-26 | 华为技术有限公司 | Load-balancing method and device |
| CN104516952B (en) * | 2014-12-12 | 2018-02-13 | 华为技术有限公司 | A kind of memory partitioning dispositions method and device |
| CN106708551B (en) * | 2015-11-17 | 2020-01-17 | 华为技术有限公司 | Configuration method and system for CPU (central processing unit) of hot-adding CPU (central processing unit) |
| CN105590063B (en) * | 2015-12-25 | 2019-03-22 | 珠海豹趣科技有限公司 | A kind of method, apparatus and electronic equipment for excavating loophole |
| CN106020971B (en) * | 2016-05-10 | 2020-01-31 | 广东睿江云计算股份有限公司 | CPU scheduling method and device in cloud host system |
| CN107577530A (en) * | 2016-07-04 | 2018-01-12 | 中兴通讯股份有限公司 | Board, the method and system of balanced board memory usage |
| CN113626214B (en) * | 2021-07-16 | 2024-02-09 | 浪潮电子信息产业股份有限公司 | Information transmission method, system, electronic equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6269390B1 (en) * | 1996-12-17 | 2001-07-31 | Ncr Corporation | Affinity scheduling of data within multi-processor computer systems |
| CN102369511A (en) * | 2011-09-01 | 2012-03-07 | 华为技术有限公司 | Resource removing method, device and system |
| CN102984762A (en) * | 2012-12-12 | 2013-03-20 | 中国联合网络通信集团有限公司 | Method and device for function allocation of IMS |
| CN103036959A (en) * | 2012-12-07 | 2013-04-10 | 武汉邮电科学研究院 | Realization method and realization system of distributed deployment application program based on input/output (IO) decoupling |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103365717B (en) * | 2013-06-25 | 2016-08-31 | 华为技术有限公司 | Memory pool access method, Apparatus and system |
-
2013
- 2013-06-25 CN CN201310257057.5A patent/CN103365717B/en active Active
-
2014
- 2014-01-23 WO PCT/CN2014/071252 patent/WO2014206078A1/en active Application Filing
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6269390B1 (en) * | 1996-12-17 | 2001-07-31 | Ncr Corporation | Affinity scheduling of data within multi-processor computer systems |
| CN102369511A (en) * | 2011-09-01 | 2012-03-07 | 华为技术有限公司 | Resource removing method, device and system |
| CN103036959A (en) * | 2012-12-07 | 2013-04-10 | 武汉邮电科学研究院 | Realization method and realization system of distributed deployment application program based on input/output (IO) decoupling |
| CN102984762A (en) * | 2012-12-12 | 2013-03-20 | 中国联合网络通信集团有限公司 | Method and device for function allocation of IMS |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103365717A (en) | 2013-10-23 |
| WO2014206078A1 (en) | 2014-12-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103365717B (en) | Memory pool access method, Apparatus and system | |
| CN109857445A (en) | Storage system and control software layout method | |
| ES2288745T3 (en) | AUTOMATION NETWORK WITH NETWORK COMPONENTS THAT GENERATE STATE MESSAGES. | |
| EP2751698B1 (en) | Computer system with processor local coherency for virtualized input/output | |
| US20110145452A1 (en) | Methods and apparatus for distribution of raid storage management over a sas domain | |
| KR102092660B1 (en) | Cpu and multi-cpu system management method | |
| JP2008517358A5 (en) | ||
| CN109886693B (en) | Consensus realization method, device, equipment and medium for block chain system | |
| JP2004021556A5 (en) | ||
| CN101894060A (en) | Fault detection method and modular device | |
| US20080120625A1 (en) | Data Processing System, Method and Interconnect Fabric that Protect Ownership Transfer with Non-Uniform Protection Windows | |
| US7734876B2 (en) | Protecting ownership transfer with non-uniform protection windows | |
| CN110971480B (en) | Computer network condition monitoring method and device, computer equipment and storage medium | |
| CN101216781A (en) | Multiprocessor system, device and method | |
| CN107624178B (en) | Fast zeroing rack architecture (RSA) and Shared Memory Controller (SMC) techniques | |
| US10592465B2 (en) | Node controller direct socket group memory access | |
| CN114039798A (en) | Data transmission method and device and electronic equipment | |
| US7904676B2 (en) | Method and system for achieving varying manners of memory access | |
| CN105353984A (en) | Floppy disk array-based high-availability cluster controller and control method and system | |
| US20130339591A1 (en) | Relaying apparatus, relay history recording method, and data processing apparatus | |
| CN106155910A (en) | A kind of methods, devices and systems realizing internal storage access | |
| US20170262209A1 (en) | Memory-driven out-of-band management | |
| CN108021476B (en) | Test method and device of interconnection interface and computing equipment | |
| CN104394232A (en) | Independent management and concentrated management method of cloud equipment | |
| CN105938461A (en) | DMA data transmission method, device and network equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |