CN107220126A - X86 servers dynamic hard partitioning method, device, storage medium and computer equipment - Google Patents
X86 servers dynamic hard partitioning method, device, storage medium and computer equipment Download PDFInfo
- Publication number
- CN107220126A CN107220126A CN201710396261.3A CN201710396261A CN107220126A CN 107220126 A CN107220126 A CN 107220126A CN 201710396261 A CN201710396261 A CN 201710396261A CN 107220126 A CN107220126 A CN 107220126A
- Authority
- CN
- China
- Prior art keywords
- subregion
- core engine
- servers
- state
- monitoring
- 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.)
- Granted
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]
- G06F9/5061—Partitioning or combining of resources
- G06F9/5077—Logical partitioning of resources; Management or configuration of virtualized resources
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/30—Monitoring
- G06F11/3051—Monitoring arrangements for monitoring the configuration of the computing system or of the computing system component, e.g. monitoring the presence of processing resources, peripherals, I/O links, software programs
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Quality & Reliability (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention relates to a kind of dynamic hard partitioning method of x86 servers, device, storage medium and computer equipment, the core engine of each node of oracle database by controlling to be deployed in x86 servers in group, partition initialization processing is carried out to corresponding node.Load monitoring is carried out to the subregion after corresponding node initialization process by core engine and obtains monitoring data；The state of correspondence subregion is estimated according to monitoring data and the management strategy of storage by core engine and obtains assessment result；Correspondence subregion is adjusted according to assessment result by core engine.Control group is combined with oracle database, data monitoring and assessment are carried out to each subregion using the core engine for being deployed in each node of oracle database, Mobile state adjustment is entered to subregion according to assessment result, realize the dynamic hard partitioning function of x86 servers, effective physical isolation ability is provided, can effective guarantee Oracle database service quality.
Description
Technical field
The present invention relates to areas of information technology, more particularly to a kind of dynamic hard partitioning method of x86 servers, device, deposit
Storage media and computer equipment.
Background technology
With progressively ripe, the continuous progress and the configuration of server based on x86 chip technologies of x86 server technologys
More and more higher, the overall process abilities of x86 servers is already close to being even more than low side unix server.
Contemporary database systems, such as oracle database provide the resource management function become more meticulous very much, Ke Yiti
For the fine granularity management to IO, concurrent computation resource, internal memory and CPU, but this resource management techniques are set up at shared bottom
On the global resource that layer operation system is managed, still contraction speed is slower for the easy Dynamic expansion of resource.And x86 servers are simultaneously
Logical partition (LPAR) such as IBM unix servers is not provided, or HPE unix servers subregion (nPar) function, use
So that the unix server of height configuration to be divided into degree of physical to isolate and run " server " of individually operated system.Based on current
X86 server hardwares function effective physical isolation ability can not be provided, to ensure Oracle database service quality.
The content of the invention
Based on this, it is necessary in view of the above-mentioned problems, provide it is a kind of can effective guarantee Oracle database service quality x86
Server dynamic hard partitioning method, device, storage medium and computer equipment.
A kind of dynamic hard partitioning method of x86 servers, comprises the following steps：
By the core engine for each node of oracle database for controlling to be deployed in x86 servers in group, to correspondence section
Point carries out partition initialization processing；
Load monitoring is carried out to the subregion after corresponding node initialization process by the core engine and obtains monitoring data；
The state of correspondence subregion is carried out according to the monitoring data and the management strategy of storage by the core engine
Assessment obtains assessment result；
Correspondence subregion is adjusted according to the assessment result by the core engine.
A kind of dynamic hard partitioning device of x86 servers, including：
Partition initialization module, for each node of oracle database by controlling to be deployed in x86 servers in group
Core engine, to corresponding node carry out partition initialization processing；
Data monitoring module, for being loaded by the core engine to the subregion after corresponding node initialization process
Monitoring obtains monitoring data；
State estimation module, for by the core engine according to the monitoring data and storage management strategy to right
Answer the state of subregion to be estimated and obtain assessment result；
Subregion adjusting module, for being adjusted by the core engine according to the assessment result to correspondence subregion.
A kind of computer-readable recording medium, is stored thereon with computer program, and the program is realized when being executed by processor
The step of above method.
A kind of computer equipment, including memory, x86 servers and store on a memory and can be in x86 servers
The computer program of upper operation, the step of x86 servers realize the above method when performing described program.
The dynamic hard partitioning method of above-mentioned x86 servers, device, storage medium and computer equipment, by controlling in group
The core engine of each node of oracle database of x86 servers is deployed in, partition initialization processing is carried out to corresponding node.It is logical
Cross core engine and monitoring data is obtained to the subregion progress load monitoring after corresponding node initialization process；Pass through core engine root
The state of correspondence subregion is estimated according to monitoring data and the management strategy of storage and obtains assessment result；Pass through core engine root
Correspondence subregion is adjusted according to assessment result.Control group is combined with oracle database, using being deployed in Oracle
The core engine of each node of database carries out data monitoring and assessment to each subregion, enters Mobile state to subregion according to assessment result and adjusts
It is whole, realize that the dynamic hard partitioning function of x86 servers, can effective guarantee Oracle data there is provided effective physical isolation ability
Storehouse service quality.
Brief description of the drawings
Fig. 1 is the flow chart of the dynamic hard partitioning method of x86 servers in an embodiment；
Fig. 2 is the flow chart of the dynamic hard partitioning method of x86 servers in another embodiment；
Fig. 3 realizes Organization Chart for dynamic hard partitioning technology in an embodiment；
Fig. 4 is the Organization Chart of core engine in an embodiment；
Fig. 5 assesses schematic diagram for state machine state in core engine in an embodiment；
Fig. 6 is subregion state Extension migratory schematic diagram in an embodiment；
Fig. 7 is low-load subregion schematic diagram in an embodiment；
Fig. 8 is intermediate part load fluctuation subregion schematic diagram in an embodiment；
Fig. 9 is subregion state contraction migration schematic diagram in an embodiment；
The schematic diagram that Figure 10 is run for subregion in an embodiment with high load condition；
The schematic diagram that Figure 11 is run for subregion in an embodiment with low-load state；
Figure 12 is actuator instruction flow schematic diagram in an embodiment；
Figure 13 is subregion load testing schematic diagram in an embodiment；
Figure 14 is the structure chart of the dynamic hard partitioning device of x86 servers in an embodiment；
Figure 15 is the structure chart of the dynamic hard partitioning device of x86 servers in another embodiment.
Embodiment
In one embodiment, a kind of dynamic hard partitioning method of x86 servers, it is adaptable to be configured with many multi-cores
Intel Xeon processors and the x86 servers for being configured with more internal memory.It can specifically be used using x86 as hardware platform
(SuSE) Linux OS and Oracle Database 12c databases, realize the dynamic hard partitioning technology of x86 servers.Such as Fig. 1 institutes
Show, this method comprises the following steps：
Step S110：By the core engine for each node of oracle database for controlling to be deployed in x86 servers in group,
Partition initialization processing is carried out to corresponding node.
Control group (Control Groups, CGroup) be linux kernel provide one kind can limit, record, every
From the mechanism of material resources (such as cpu memory i/o) used in process group (process groups).Specifically,
CGroup is the linux kernel function that any process is carried out to packetizing management, itself is to provide process carrying out packetizing
The function of management and the foundation structure of interface, the specific resource management function such as the distribution control of I/O or internal memory is by this
Function is realized.These specific resource management functions are referred to as CGroup subsystems or controller.CGroup subsystems have control
The Memory controllers of internal memory processed, the cpu controller for controlling process scheduling etc..CGroup that operating kernel can be used
System is confirmed by/proc/cgroup.CGroup provides a CGroup Virtual File System as progress grouping management
The user interface set with each subsystem.The necessary carry CGroup file system of CGroup is used, at this moment passes through carry option
Specify and which subsystem used.The kind of document that CGroup is supported is as shown in table 1.
Table 1
In CGroup, task is exactly a process of system.One group of group is controlled to be divided according to certain standard
Process.Resources control in CGroup is realized in units of controlling group.One process can be added to some control race
Group, also moves to another control group from a process group.The process of one process group can use CGroup to control race
The resource that group distributes for unit, is limited while being set by CGroup in units of controlling group.
Control group can be organized into the control group tree of level (hierarchy) form, i.e., one.Control on group tree
Child node control group be father node control group child, inherit father control group specific attribute.One subsystem
It is exactly a resouce controller, such as cpu subsystem is exactly a controller for controlling the CPU time to distribute.Subsystem must be attached
Being added in a level can just be worked, and a subsystem is attached to after some level, all control races in this level
Group is controlled by this subsystem.When creating new level in systems every time, all tasks in the system are all those
Level acquiescence CGroup (root CGroup can be referred to as, this CGroup create level when automatically create, behind in the layer
The CGroup that creates is this CGroup offspring in level) initial member.One subsystem can only at most be attached to a layer
Level a, level can add multiple subsystems.One task can be multiple CGroup member, but these CGroup must
Must be in different levels.When process (task) in system creates subprocess (task), the subtask automatically becomes its parent process
Place CGroup member.Then the subtask can be moved in different CGroup as needed, but start when it always
Inherit the CGroup of his father's task.
In the present embodiment, oracle database uses oracle database 12c, introduces processor_group
(Oracle, Oracle Database 12c Release 1 (12.1.0.1) New Features, 2016) characteristic, utilizes this
One characteristic, oracle database 12c can combine closely with Linux or Solaris, so as to utilize the processor of operating system
Set and its correlation resource, realize by database instance process and internal memory be tied to the specific processor of operating system aspect and
On its memory source, and this characteristic is primarily limited to the NUMA (Non-uniform that current x86 server designs are used
Memory access, nonuniform memory access framework) framework.NUMA is a kind of internal memory of the computer design for multiprocessor, interior
Deposit position of the access time depending on internal memory relative to processor.Under NUMA, processor accesses the local memory of its own
Speed is faster than non-local internal memory (internal memory is located at the internal memory shared between another processor, or processor).NUMA
The characteristics of be：The internal memory being shared is physically distributed, and the set of all these internal memories is exactly global address space.So
The time that processor accesses these internal memories is different, it is clear that access the speed of local memory than accessing global sharing memory
Or the nonlocal internal memory of remote access is quicker.In addition, internal memory is probably layering in NUMA：Shared drive in local memory, group, entirely
Office's shared drive.
Based on current Intel Xeon processor technologies, current 4 road server possible configurations 64 cores, 128 lines
Journey；If two-way server possible configuration 32 cores, 64 threads.For many medium-sized and small enterprises, if such flat
Single database is run on platform, it will usually which utilization rate is not enough, the integration for carrying out business is just imperative.Current 8 tunnels service
The core number of device may be 192 cores, 384 threads.Comprehensive 4 common tunnels, from the point of view of 8 road server configurations, current hardware handles
Ability maintains an equal level with many unix servers in performance, or even has surmounted unix server.But surmounting in performance can not be
Enterprise brings stable running environment, and possesses very strong practicality using extensive partitioning technique on traditional unix server,
Because this technology can provide necessary isolation, so as to protect on the basis of hardware resource is made full use of for enterprise's application
Demonstrate,prove the stability of enterprise's application.
It is real during by running (SuSE) Linux OS and oracle database 12c progress business integrations on x86 servers
Show the dynamic hard partitioning of database instance to ensure database service quality.Dynamic hard partitioning, which refers to, is deployed in same x86pc clothes
The database instance being engaged on device monopolizes a data CPU, or one in a CPU or several calculating cores, works as consumption
To the greatest extent its exclusive computing resource when, the database instance will not take the computing resource that other database instances are distributed so that
Ensure the database service quality on the x86 servers.
Specifically, by each node for controlling the core engine in group to be arranged in oracle database, core engine
Quantity is corresponding with number of nodes.Core engine belongs to micro-kernel product, i.e., all components are interior to be built in core engine, component
Between close coupling, finally ensure the stabilization, reliable of this core engine.Core engine is responsible for book server computer resource controller
Initialization, it is ensured that availability of each computing resource during running.
Step S130：Load monitoring is carried out by core engine to the subregion after corresponding node initialization process to be monitored
Data.
The load of all subregions in present node is monitored using core engine, after the monitoring data of acquisition is used as
The continuous foundation for carrying out subregion adjustment.
Step S140：The state of correspondence subregion is carried out according to monitoring data and the management strategy of storage by core engine
Assessment obtains assessment result.
Specifically management strategy can be received in advance to be preserved, plan that can be with the new management strategy of real-time reception to having stored
Slightly it is updated.The species of management strategy is not unique, divides from the scope of application, and management strategy may include Unified Policy and independent
Strategy.Wherein, Unified Policy refers to all general, the tactful each change of all nodes of the strategy in cluster, it is necessary to be copied to
Whole cluster, it is ensured that each subregion (or database instance) possesses consistent computing resource.Independent strategy refers to the strategy and only existed
On some single server effectively, the strategy need not be copied to whole cluster.Divide from purposes, management strategy may include extension
Strategy and reduced technique.Wherein, if the computing resource that expanding policy refers in current bay reaches default upper limit threshold value,
Stable operation generally for business usually requires to be extended the computing resource of current bay, and the extension of computing resource is usual
It can be migrated according to upper limit threshold value.If the computing resource that reduced technique refers in current bay is in idle state, generally
Need to shrink the computing resource of current bay when laying in computing resource for other subregions of present node, computing resource
Contraction would generally be shunk according to lower limit threshold value.Reduced technique is not the simple contrary operation of expanding policy, shrinks plan
The release by stopping database instance progress resource is slightly needed, resource then could be discharged.
Extension or contraction are determined by core engine combination monitoring data and obtain assessment result, are then tied according to assessment
Fruit carries out final extension or contraction.In one embodiment, management strategy includes subregion state pass corresponding with threshold value
System, step S140 includes step 142 and step 144.
Step 142：According to monitoring data and subregion state and the corresponding relation of threshold value, the target-like of correspondence subregion is obtained
State.
The threshold value formulated dependent on policy manager is shunk in the extension of subregion computing resource, the setting of threshold value generally according to
Rely the hardware configuration and O＆M monitoring criteria in present node.Dbjective state is to refer to the state that subregion needs to migrate.Illustrate
Bright, subregion state includes tri- kinds of A, B and C, and corresponding threshold value is respectively a, b and c.If the monitoring data obtained and threshold value c
Matching, then the dbjective state of the subregion is C.
Step 144：The assessment result of correspondence subregion is obtained according to the current state and dbjective state of each subregion respectively.
Exemplified by the same dbjective state by subregion is C, if the current state of subregion is A, obtained assessment result is A->
C is, it is necessary to which the subregion is migrated to C-state from A condition.
Step S150：Enter Mobile state to correspondence subregion according to assessment result by core engine to adjust.
After the assessment result of each subregion is obtained, core engine directly enters Mobile state according to assessment result to correspondence subregion
Adjustment.Specifically, it may include that expanding policy and reduced technique are corresponding with management strategy, step S150 includes：According to assessment result
Computing resource to correspondence subregion expands or contracts adjustment.
In one embodiment, as shown in Fig. 2 after step S110, before step S130, this method also includes step
S120。
Step S120：Isolation test is carried out to the subregion after initialization process by core engine.
Subregion is after configuration is finished, it usually needs carry out the isolation test of each subregion, can be single in each subregion
Following script is solely run, to verify the high degree of isolation of subregion, the dynamic hard partitioning reliability of x86 servers is improved.Utilize core
Heart engine carries out isolation test, and the step S130 carried out after isolation is tested and passed through to the subregion after initialization process.
In one embodiment, after step S140, in addition to by controlling the messaging bus in group to carry out each core
The step of message between engine is transmitted.
Group is controlled also to include the messaging bus for connecting each core engine, the message that messaging bus is responsible between each node is passed
Pass, message transmission specifically may include the transmission to Unified Policy, it is ensured that uniformity of the strategy in cluster.Message transmission can also be wrapped
Include the transmission instructed to zoned migration, it is ensured that the uniformity of the subregion state of other nodes in cluster.If it is appreciated that being
System only has a node, configures single core engine, it is not necessary to dispose messaging bus；Only when disposing cluster environment
Just need the message transmission disposed between messaging bus progress core engine.
For the ease of more fully understanding the dynamic hard partitioning method of above-mentioned x86 servers, enter with reference to specific embodiment
The detailed explanation of row.
A weight of the realization of dynamic hard partitioning dependent on the CGroup and oracle database 12c of (SuSE) Linux OS
Characteristic Processor_Group_Name is wanted, similar UNIX services can be realized on x86 servers when the two is combined
The dynamic hard partitioning function of just possessing on device, dynamic hard partitioning technology realizes Organization Chart as indicated at 3.Dispose on each node dynamic
The transmission of message is carried out between the core engine of state hard partitioning, core engine by the messaging bus outside node.X86 servers
Dynamic hard partitioning is driven and managed by core engine, and its Organization Chart is as shown in Figure 4.Core engine includes monitor, state machine, plan
Omit manager and actuator.Monitor is responsible for the load monitoring of all subregions in present node, and state machine combines the plan of configuration
The assessment of each subregion state is slightly carried out, and assessment of the execution state machine to each subregion capacity finally is finally responsible for by actuator
As a result；Policy manager is responsible for the duplication in whole cluster of configuration and strategy of strategy, it is ensured that strategy in cluster one
Cause property.The core engine of each node is all in the status of equality, and the operation that each engine is initiated can broadcast or replicate
To the engine of other nodes.
CGroup is using preceding needing dependence cpuset subsystems normally to run, and core engine is responsible for the initial of the subsystem
Change, specific script is as follows：
#mkdir/etc/cpuset
#touch/etc/cpuset.conf
#mkdir/dev/cpuset
#mount–t cpuset cpuset/dev/cpuset
#load_partition
#service cgconfig start
#service cgred start
Wherein load_partition is responsible for reading the correct configuration of subregion out of/etc/cpuset.conf, and is entered
The suitable initialization of row, otherwise follow-up database instance can not monopolize cpu resources.
The internal memory that core engine is responsible in computing resource set is in exclusive state, and it is not transportable to ensure that internal memory is in
State, has ensured that the stability and exclusiveness of database instance running status, and specific script is as follows：
#echo 1>/dev/cpuset/PARTITION/mem_exclusive
#echo 1>/dev/cpuset/PARTITION/memory_migrate
After partition initialization is finished, database instance be can a kind of exclusive state monopolize computing resource (CPU and interior
Deposit), it can not also apply for the resource of other subregions to the resource run out even if the example in the subregion, and specific script is as follows：
SQL>Alter system set processor_group_name=ORACLE scope=spfile；
SQL>shutdown immediate
SQL>startup
Core engine is not introduced into other third party's business softwares, therefore does not increase extra cost of investment.
The load that monitor is responsible for each subregion is monitored and stored.Load monitoring is carried using system default
System monitoring instrument sar；Multiple subregions are usually there will be on same node, monitored partition is obtained first during monitoring load
Cpuset values, be then monitored, finally will collection load data storage in local workload storehouse.
If system does not configure system activity monitoring collector, self-defined monitoring collector is needed.Monitor is supervised
Control is the partition information that the first step needs to define in acquisition system, can be obtained by following script：
#！/bin/bash
cd/cgroup/cpuset
find./-type d|egrep-o"[[:alnum:]]"*
Then the cpu information in subregion is obtained：
#！/bin/bash
cgget-g cpuset:$1|grep cpuset.cpus|cut-d':'–f2
# note obtain result be a cpu core initial value, it is necessary to further processing
Further monitored according to the cpuset information of acquisition.For example as shown in following order：
#sar-P 0,15
If system configuration system activity monitors collector, monitor can regularly be carried out from sar monitored results
Analysis obtains related result and also may be used.
Policy manager is responsible for the management of monitoring strategies, and partitioned computer resource extension reduced technique management and
Replicate, in view of there may be the integration of different Service Databases in a cluster, i.e., each node is negative in same cluster
Carrying may be not consistent, therefore policy manager will be treated with a certain discrimination to different strategies.
The threshold value formulated dependent on policy manager is shunk in the extension of subregion computing resource, the setting of threshold value generally according to
Rely the hardware configuration and O＆M monitoring criteria in present node.
As shown in table 2, the subregion state migration of sample setting and common O＆M monitor control index relation table, threshold value are pressed
Amplitude according to 25% is incremented by or successively decreased.Reach after default threshold value, it usually needs according to certain quantity to calculating
Core is extended or shunk.
| State | Threshold value 1 | Threshold value 2 | Threshold value 3 |
| S1 | ≤ 25% | ||
| S2 | ≤ 50% | ||
| S3 | ≤ 75% |
Table 2
As shown in table 3, the subregion extension of sample setting is extended according to baseline progression strategy, mainly in view of dividing
Area carries out meeting being actually needed for business load during great-leap-forward migration.
| State | Extension 1 | Extension 2 | Extension 3 |
| S1 | +2 | ||
| S2 | +4 | ||
| S3 | +6 |
Table 3
As shown in table 4 relation table is shunk for the subregion that a sample is set.
| State | Shrink 1 | Shrink 2 | Shrink 3 |
| S1 | -2 | ||
| S2 | -4 | ||
| S3 | -6 |
Table 4
It is a complicated process that subregion, which shrinks, because computing resource is exclusive, it is impossible to by simply reducing computing resource
Quantity and allow subregion carry out it is dynamic shrink steadily, reliable and stable method is discharged by closing subregion (database instance)
Resource, then shrink partitions computing resource, then restarts the subregion.This is also that current virtualization container computing resource is shunk
Main contraction mode.
Pass through the computing resource extension or contraction of the effective stability contorting subregion of finite state machine, it is ensured that partition running is stable
It is controllable.The strategy and monitoring data that state machine is responsible for being formulated according to policy manager are adjusted to subregion, specific adjustment
Instruction is responsible for execution by actuator, and is delivered on each node in whole cluster.Fig. 5, which show state machine state and assessed, to be shown
It is intended to.Migration is shunk to subregion state Extension migratory and subregion state separately below to be explained.
Fig. 6 is subregion state Extension migratory schematic diagram.Fig. 7 show the Monitoring Data of a low-load subregion, the subregion one
Directly operate steadily, threshold value is extended based on the subregion that policy manager is set, the subregion need not carry out the extension of computing resource.Should
The state transition path of subregion is：S1->S1.Fig. 8 show the Monitoring Data that an intermediate part load fluctuates subregion, subregion load
Fluctuation is more obvious, extends threshold value based on the subregion that policy manager is set, it should the expansion of computing resource is carried out for the subregion
Exhibition, increases the computing resource of 2 cores.The transition state Lu Jingwei of the subregion：S1->S2.
The strategy that state machine can be formulated according to policy manager, assesses whether current bay has available resource to be expanded
Exhibition, if available resource is extended, then state machine can generate the specific instruction of computing resource extension to actuator：
echo min-current+2>/dev/cpuset/PARTITION/cpus
If be extended without available resource, state machine can send warning information to actuator.Current bay is calculated
The extension of resource can be rapidly completed, and subregion extended instruction can also be delivered to whole cluster at a terrific speed and quick right
Each subregion carries out the extension of computing resource.
Fig. 9 is that subregion state shrinks migration schematic diagram.If the resource utilization of subregion is not enough, have greatly changed,
It is generally necessary to carry out the recovery of partitioned resources.The Monitoring Data figure run as shown in Figure 10 for a subregion with high load condition, should
Subregion is illustrated in fig. 11 shown below always with higher load running if load drops to, and the subregion needs to carry out partitioned resources
Shrink.One takes complicated process when resource is shunk, and directly can not reclaim resource out of subregion, it usually needs by data
Storehouse example is shut down, then Resource recovery；Then similar operations are performed in other nodes again.
$srvctl stop instance–db orcl–instance orcl1
#echo min-current-6>/dev/cpuset/PARTITION/cpus
Actuator is that instruction final in core engine performs body, and it is responsible for performing the monitoring instruction that monitor is issued, and holds
The strategy change synchronic command that row policy manager is issued, performs the zoned migration instruction that state machine is issued, performs other execution
Device is delivered to the instruction of messaging bus, is also responsible for actuator and also loads that unified instruction is delivered into messaging bus, it is ensured that in cluster
Other nodes subregion state uniformity.It is actuator instruction flow chart as shown in figure 12.Messaging bus is loaded each
The message transmission of node.
Subregion is after configuration is finished, it usually needs carry out the isolation test of each subregion, can be single in each subregion
Following script is solely run, to verify the high degree of isolation of subregion.Specific script is as follows：
To ensure zero load running status at other subregions when running the script.If surveyed in the case where meeting condition
Examination, other subregions have higher load running status, then illustrate that the configuration of the lower layer resources such as CGroup or cpuset has mistake.Just
In the case of often, subregion load diagram is as shown in figure 13.
The dynamic hard partitioning method of above-mentioned x86 servers, control group is combined with oracle database, using being deployed in
The core engine of each node of oracle database carries out data monitoring and assessment to each subregion, and subregion is carried out according to assessment result
Dynamic adjustment, realizes that the dynamic hard partitioning function of x86 servers, can effective guarantee there is provided effective physical isolation ability
Oracle database service quality.
In one embodiment, the dynamic hard partitioning device of a kind of x86 servers, it is adaptable to be configured with many multi-cores
Intel Xeon processors and the x86 servers for being configured with more internal memory.As shown in figure 14, the device includes partition initialization
Module 110, data monitoring module 130, state estimation module 140 and subregion adjusting module 150.
Partition initialization module 110 is used for by controlling the oracle database for being deployed in x86 servers in group respectively to save
The core engine of point, partition initialization processing is carried out to corresponding node.
Specifically, by each node for controlling the core engine in group to be arranged in oracle database, core engine
Quantity is corresponding with number of nodes.Core engine belongs to micro-kernel product, i.e., all components are interior to be built in core engine, component
Between close coupling, finally ensure the stabilization, reliable of this core engine.Core engine is responsible for book server computer resource controller
Initialization, it is ensured that availability of each computing resource during running.
Data monitoring module 130 is used to carry out load prison to the subregion after corresponding node initialization process by core engine
Control obtains monitoring data.
The load of all subregions in present node is monitored using core engine, after the monitoring data of acquisition is used as
The continuous foundation for carrying out subregion adjustment.
State estimation module 140 is used for by core engine according to monitoring data and the management strategy of storage to correspondence subregion
State be estimated and obtain assessment result.
Specifically management strategy can be received in advance to be preserved, plan that can be with the new management strategy of real-time reception to having stored
Slightly it is updated.In one embodiment, management strategy includes subregion state and the corresponding relation of threshold value, state estimation module
140 include dbjective state acquiring unit and assessment result acquiring unit.
Dbjective state acquiring unit is used for the corresponding relation according to monitoring data and subregion state and threshold value, obtains correspondence
The dbjective state of subregion.
The threshold value formulated dependent on policy manager is shunk in the extension of subregion computing resource, the setting of threshold value generally according to
Rely the hardware configuration and O＆M monitoring criteria in present node.Dbjective state is to refer to the state that subregion needs to migrate.Illustrate
Bright, subregion state includes tri- kinds of A, B and C, and corresponding threshold value is respectively a, b and c.If the monitoring data obtained and threshold value c
Matching, then the dbjective state of the subregion is C.
Assessment result acquiring unit is used to obtain correspondence subregion according to the current state and dbjective state of each subregion respectively
Assessment result.
Exemplified by the same dbjective state by subregion is C, if the current state of subregion is A, obtained assessment result is A->
C is, it is necessary to which the subregion is migrated to C-state from A condition.
Subregion adjusting module 150 is used to enter Mobile state adjustment to correspondence subregion according to assessment result by core engine.
After the assessment result of each subregion is obtained, core engine directly enters Mobile state according to assessment result to correspondence subregion
Adjustment.Specifically, it may include that expanding policy and reduced technique are corresponding with management strategy, subregion adjusting module 150 is with specific reference to commenting
Estimate result and adjustment is expanded or contracted to the computing resource of correspondence subregion.
In one embodiment, as shown in figure 15, the dynamic hard partitioning device of x86 servers also includes isolation test module
120。
Isolation test module 120 is used to carry out partition initialization processing to corresponding node in partition initialization module 110
Afterwards, data monitoring module 130 is obtained by core engine to the subregion progress load monitoring after corresponding node initialization process
Before monitoring data, isolation test is carried out to the subregion after initialization process by core engine, and it is logical in isolation test
Later control data monitoring module 130 is obtained by core engine to the subregion progress load monitoring after corresponding node initialization process
To monitoring data.
In one embodiment, the dynamic hard partitioning device of x86 servers also includes message transmission module.
Message transmission module is used in state estimation module 140 by core engine according to monitoring data and the management of storage
Strategy is estimated to the state of correspondence subregion obtain assessment result after, by controlling the messaging bus in group to carry out each core
Message transmission between heart engine.
Group is controlled also to include the messaging bus for connecting each core engine, the message that messaging bus is responsible between each node is passed
Pass, message transmission specifically may include the transmission to Unified Policy, it is ensured that uniformity of the strategy in cluster.Message transmission can also be wrapped
Include the transmission instructed to zoned migration, it is ensured that the uniformity of the subregion state of other nodes in cluster.If it is appreciated that being
System only has a node, configures single core engine, it is not necessary to dispose messaging bus；Only when disposing cluster environment
Need the message transmission disposed between messaging bus progress core engine.
The dynamic hard partitioning device of above-mentioned x86 servers, control group is combined with oracle database, using being deployed in
The core engine of each node of oracle database carries out data monitoring and assessment to each subregion, and subregion is carried out according to assessment result
Dynamic adjustment, realizes that the dynamic hard partitioning function of x86 servers, can effective guarantee there is provided effective physical isolation ability
Oracle database service quality.
In one embodiment, a kind of computer-readable recording medium, is stored thereon with computer program, and the program is located
The step of reason device realizes the above method when performing.Storage medium can be specifically floppy disk, CD, DVD, hard disk, flash memory, USB flash disk etc.,
Particular type is not unique.
Above computer readable storage medium storing program for executing, control group is combined with oracle database, using being deployed in
The core engine of each node of oracle database carries out data monitoring and assessment to each subregion, and subregion is carried out according to assessment result
Dynamic adjustment, realizes that the dynamic hard partitioning function of x86 servers, can effective guarantee there is provided effective physical isolation ability
Oracle database service quality.
In one embodiment, a kind of computer equipment, including memory, x86 servers and storage are on a memory
And the computer program that can be run on x86 servers, the step of realizing the above method during x86 server configuration processors.
Above computer equipment, control group is combined with oracle database, using being deployed in oracle database
The core engine of each node carries out data monitoring and assessment to each subregion, and Mobile state adjustment is entered to subregion according to assessment result, real
The dynamic hard partitioning function of existing x86 servers, can effective guarantee Oracle database service there is provided effective physical isolation ability
Quality.
Each technical characteristic of embodiment described above can be combined arbitrarily, to make description succinct, not to above-mentioned reality
Apply all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, the scope of this specification record is all considered to be.
Embodiment described above only expresses the several embodiments of the present invention, and it describes more specific and detailed, but simultaneously
Can not therefore it be construed as limiting the scope of the patent.It should be pointed out that coming for one of ordinary skill in the art
Say, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to the protection of the present invention
Scope.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (10)
1. a kind of dynamic hard partitioning method of x86 servers, it is characterised in that comprise the following steps：
By the core engine for each node of oracle database for controlling to be deployed in x86 servers in group, corresponding node is entered
The processing of row partition initialization；
Load monitoring is carried out to the subregion after corresponding node initialization process by the core engine and obtains monitoring data；
The state of correspondence subregion is estimated according to the monitoring data and the management strategy of storage by the core engine
Obtain assessment result；
Enter Mobile state to correspondence subregion according to the assessment result by the core engine to adjust.
2. the dynamic hard partitioning method of x86 servers according to claim 1, it is characterised in that described by controlling group
In be deployed in x86 servers each node of oracle database core engine, to corresponding node carry out partition initialization processing
Afterwards, it is described that monitoring data is obtained to the subregion progress load monitoring after corresponding node initialization process by the core engine
Before, it is further comprising the steps of：
Isolation test is carried out to the subregion after initialization process by the core engine, and tested in isolation by laggard
Row is described to obtain monitoring data by the core engine to the subregion progress load monitoring after corresponding node initialization process
Step.
3. the dynamic hard partitioning method of x86 servers according to claim 1, it is characterised in that the management strategy includes
The corresponding relation of subregion state and threshold value, it is described by the core engine according to the monitoring data and default management plan
Slightly the state for corresponding to subregion is estimated and obtains assessment result, including：
According to the monitoring data and subregion state and the corresponding relation of threshold value, the dbjective state of correspondence subregion is obtained；
The assessment result of correspondence subregion is obtained according to the current state and dbjective state of each subregion respectively.
4. the dynamic hard partitioning method of x86 servers according to claim 3, it is characterised in that described to pass through the core
Engine enters Mobile state adjustment to correspondence subregion according to the assessment result, including：According to the assessment result to correspondence subregion
Computing resource expands or contracts adjustment.
5. the dynamic hard partitioning method of x86 servers according to claim 1, it is characterised in that described to pass through the core
Engine is estimated according to the monitoring data and the management strategy of storage to the state of correspondence subregion obtain assessment result after,
It is further comprising the steps of：
The message between each core engine is carried out by the messaging bus in the control group to transmit.
6. a kind of dynamic hard partitioning device of x86 servers, it is characterised in that including：
Partition initialization module, the core for each node of oracle database by controlling to be deployed in x86 servers in group
Heart engine, partition initialization processing is carried out to corresponding node；
Data monitoring module, for carrying out load monitoring to the subregion after corresponding node initialization process by the core engine
Obtain monitoring data；
State estimation module, for being divided by the core engine according to the monitoring data and the management strategy of storage correspondence
The state in area, which is estimated, obtains assessment result；
Subregion adjusting module, is adjusted for entering Mobile state to correspondence subregion according to the assessment result by the core engine.
7. the dynamic hard partitioning device of x86 servers according to claim 6, it is characterised in that also include：
Isolation test module, after partition initialization processing is carried out to corresponding node in the partition initialization module,
The data monitoring module carries out load monitoring to the subregion after corresponding node initialization process by the core engine and obtained
Before monitoring data, isolation test is carried out to the subregion after initialization process by the core engine, and survey in isolation
Ping and control the data monitoring module to be carried out by the core engine to the subregion after corresponding node initialization process later
Load monitoring obtains monitoring data.
8. the dynamic hard partitioning device of x86 servers according to claim 6, it is characterised in that also include：
Message transmission module, in the state estimation module by the core engine according to the monitoring data and storage
Management strategy the state of correspondence subregion be estimated obtain assessment result after, it is total by the message in the control group
Line carries out the message transmission between each core engine.
9. a kind of computer-readable recording medium, is stored thereon with computer program, it is characterised in that the program is held by processor
The step of any one methods described in claim 1-5 is realized during row.
10. a kind of computer equipment, including memory, x86 servers and store on a memory and can be on x86 servers
The computer program of operation, it is characterised in that the x86 servers are realized any in claim 1-5 when performing described program
The step of one methods described.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710396261.3A CN107220126B (en) | 2017-05-27 | 2017-05-27 | X86 server dynamic hard partition method, device, storage medium and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710396261.3A CN107220126B (en) | 2017-05-27 | 2017-05-27 | X86 server dynamic hard partition method, device, storage medium and computer equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107220126A true CN107220126A (en) | 2017-09-29 |
| CN107220126B CN107220126B (en) | 2020-12-01 |
Family
ID=59947544
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710396261.3A Active CN107220126B (en) | 2017-05-27 | 2017-05-27 | X86 server dynamic hard partition method, device, storage medium and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107220126B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112380108A (en) * | 2020-07-10 | 2021-02-19 | 中国航空工业集团公司西安飞行自动控制研究所 | Full-automatic test method for partition space isolation |
| CN113835845A (en) * | 2021-11-26 | 2021-12-24 | 长沙驭电信息技术有限公司 | Method and system for realizing hard partition capacity of memory bound by CPU core |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060101224A1 (en) * | 2004-11-08 | 2006-05-11 | Shah Punit B | Autonomic self-tuning of database management system in dynamic logical partitioning environment |
| US20120030686A1 (en) * | 2010-07-29 | 2012-02-02 | International Business Machines Corporation | Thermal load management in a partitioned virtual computer system environment through monitoring of ambient temperatures of envirnoment surrounding the systems |
| CN103365923A (en) * | 2012-03-30 | 2013-10-23 | 伊姆西公司 | Method and device for assessing partition schemes of database |
| CN104333488A (en) * | 2014-11-04 | 2015-02-04 | 哈尔滨工业大学 | Cloud service platform performance test method |
| CN105183565A (en) * | 2015-09-30 | 2015-12-23 | 华为技术有限公司 | Computer and service quality control method and device |
| CN105608138A (en) * | 2015-12-18 | 2016-05-25 | 贵州大学 | System for optimizing parallel data loading performance of array databases |
-
2017
- 2017-05-27 CN CN201710396261.3A patent/CN107220126B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060101224A1 (en) * | 2004-11-08 | 2006-05-11 | Shah Punit B | Autonomic self-tuning of database management system in dynamic logical partitioning environment |
| US20120030686A1 (en) * | 2010-07-29 | 2012-02-02 | International Business Machines Corporation | Thermal load management in a partitioned virtual computer system environment through monitoring of ambient temperatures of envirnoment surrounding the systems |
| CN103365923A (en) * | 2012-03-30 | 2013-10-23 | 伊姆西公司 | Method and device for assessing partition schemes of database |
| CN104333488A (en) * | 2014-11-04 | 2015-02-04 | 哈尔滨工业大学 | Cloud service platform performance test method |
| CN105183565A (en) * | 2015-09-30 | 2015-12-23 | 华为技术有限公司 | Computer and service quality control method and device |
| CN105608138A (en) * | 2015-12-18 | 2016-05-25 | 贵州大学 | System for optimizing parallel data loading performance of array databases |
Non-Patent Citations (4)
| Title |
|---|
| 张建华，等主编；: "《计算机网络基础 2013年1月第1版》", 31 January 2013, 北京邮电大学出版社 * |
| 李爱武，编著，: "《融会贯通，从Oracle 12C到SQL Server 2016 2016年12月第1版》", 31 December 2016 * |
| 赵炎，等；: "针对云计算IaaS隔离性的测试系统", 《计算机系统应用》 * |
| 赵科，: "数据库联机事务处理测评系统中ACID子系统的实现和InnoDB事务处理分析", 《中国优秀博硕士学位论文全文数据库（硕士）信息科技辑（月刊）》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112380108A (en) * | 2020-07-10 | 2021-02-19 | 中国航空工业集团公司西安飞行自动控制研究所 | Full-automatic test method for partition space isolation |
| CN112380108B (en) * | 2020-07-10 | 2023-03-14 | 中国航空工业集团公司西安飞行自动控制研究所 | Full-automatic test method for partition space isolation |
| CN113835845A (en) * | 2021-11-26 | 2021-12-24 | 长沙驭电信息技术有限公司 | Method and system for realizing hard partition capacity of memory bound by CPU core |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107220126B (en) | 2020-12-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9977689B2 (en) | Dynamic scaling of management infrastructure in virtual environments | |
| US11226750B2 (en) | Flexible deprovisioning of distributed storage | |
| US11847329B2 (en) | Techniques for implementing fault domain sets | |
| Li et al. | Tachyon: Reliable, memory speed storage for cluster computing frameworks | |
| US9672116B1 (en) | Backup using instinctive preferred server order list (PSOL) | |
| Hoffmann et al. | Megaphone: Latency-conscious state migration for distributed streaming dataflows | |
| US20240061812A1 (en) | Metadata control in a load-balanced distributed storage system | |
| US11409449B2 (en) | Workload based storage optimization | |
| US10909072B2 (en) | Key value store snapshot in a distributed memory object architecture | |
| Zhang et al. | Improving Hadoop service provisioning in a geographically distributed cloud | |
| Ma et al. | Dependency-aware data locality for MapReduce | |
| CN112714018A (en) | Gateway-based ElasticSearch search service method, system, medium and terminal | |
| Premchaiswadi et al. | Optimizing and tuning MapReduce jobs to improve the large‐scale data analysis process | |
| CN107220126A (en) | X86 servers dynamic hard partitioning method, device, storage medium and computer equipment | |
| US10969959B2 (en) | Effective resynchronization in virtual storage area network using dynamic partitioning and data deduplication techniques | |
| US10140021B2 (en) | Adaptive data-partitioning model that responds to observed workload | |
| Kumar et al. | Graphsteal: Dynamic re-partitioning for efficient graph processing in heterogeneous clusters | |
| US10067949B1 (en) | Acquired namespace metadata service for controlling access to distributed file system | |
| US20150212847A1 (en) | Apparatus and method for managing cache of virtual machine image file | |
| EP3519985A1 (en) | Flexible in-memory column store placement | |
| GB2585543A (en) | Data migration in a hierarchical storage management system | |
| JP2014153935A (en) | Parallel distributed processing control device, parallel distributed processing control system, parallel distributed processing control method, and parallel distributed processing control program | |
| Salapura et al. | Enabling enterprise-class workloads in the cloud | |
| Chawla | Optimizing the Resource utilization of Enterprise Content management workloads through measured performance baselines and dynamic topology adaptation | |
| KR20150094181A (en) | Method for control of speculative execution based on node performance and device therefor |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200610 Address after: Room 208, No. 100, Dongxing Road, Donghuan street, Panyu District, Guangzhou City, Guangdong Province Applicant after: SOUTHERN POWER GRID PEAK LOAD AND FREQUENCY REGULATION POWER GENERATION Co.,Ltd. Address before: 510635 No. 32 Longkou East Road, Guangzhou, Guangdong, Tianhe District Applicant before: China Southern Power Grid Tiaofeng Frequency Modulation Power Generation Company |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230830 Address after: Room 1503, No. 858, Lianhua Avenue West, Donghuan Street, Panyu District, Guangzhou, Guangdong 510000 Patentee after: Southern Power Grid Energy Storage Co.,Ltd. Information and Communication Branch Address before: 511400 room 208, 100 Dongxing Road, Donghuan street, Panyu District, Guangzhou City, Guangdong Province Patentee before: SOUTHERN POWER GRID PEAK LOAD AND FREQUENCY REGULATION POWER GENERATION Co.,Ltd. |