CN106528488A - Computation cluster system in single machine and control method - Google Patents
Computation cluster system in single machine and control method Download PDFInfo
- Publication number
- CN106528488A CN106528488A CN201610883006.7A CN201610883006A CN106528488A CN 106528488 A CN106528488 A CN 106528488A CN 201610883006 A CN201610883006 A CN 201610883006A CN 106528488 A CN106528488 A CN 106528488A
- Authority
- CN
- China
- Prior art keywords
- cpu
- distribution
- core
- network interface
- services module
- 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
- G06F15/00—Digital computers in general; Data processing equipment in general
- G06F15/16—Combinations of two or more digital computers each having at least an arithmetic unit, a program unit and a register, e.g. for a simultaneous processing of several programs
- G06F15/161—Computing infrastructure, e.g. computer clusters, blade chassis or hardware partitioning
Landscapes
- Engineering & Computer Science (AREA)
- Computer Hardware Design (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Multi Processors (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
- Computer And Data Communications (AREA)
Abstract
The invention discloses a computation cluster system in a single machine and a control method. The system comprises a service distribution module, multiple network cards, a CPU and multiple hard disks, wherein the CPU comprises multiple computation cores; the multiple network cards are in one-to-one correspondence to the multiple computation cores of the CPU, so multiple computation nodes in the single machine are constituted; and the CPU is connected to the multiple hard disks by data buses. According to the invention, the multiple computation cores and the multiple network cards are set as the multiple computation nodes in the single machine, and are connected and bound to the multiple disks; application requests can be distributed rationally by the service distribution module; cluster computation with high IO performance and high efficiency in the single machine can be achieved at the low cost; and the system is perfectly adapted to multiple access networks. The computation cluster system in the single machine and the control method disclosed by the invention can be widely applied in the field of data processing.
Description
Technical field
The present invention relates to computing cluster system and control method in data processing field, especially a kind of unit.
Background technology
Current all of high-performance, high IO computing clusters are all based on multiple stage physical server, a server conduct
One calculate node, so that form computing cluster.Due to relying on multiple stage physical server, cost is relatively high.
In order to solve the above problems, in prior art in a kind of cluster unit multinode implementation method
(CN102999571B)By disposing multiple nodes on a server, without the need for relying on multiple stage physical server, but its side again
Method only illustrates multiple nodes and server disk of deployment, is accomplished that the shared of all nodal informations in cluster, solves
The limitation of polylith disk cannot be made full use of, but and increases disposal ability;Especially in the application scenario of mass data IO,
Miscellaneous due to reading and writing that data volume is big, species many, server is between a variety of data, frequently cut between read-write operation
Change, cause that treatment effeciency is low, error rate is high.
The content of the invention
In order to solve above-mentioned technical problem, the purpose of the present invention is:There is provided and in a kind of unit, realize that high-performance, high IO are fast
The computing cluster system of degree.
In order to solve above-mentioned technical problem, it is another object of the present invention to:There is provided and in a kind of unit, realize high-performance, height
The computing cluster control method of IO speed.
The technical solution adopted in the present invention is:Computing cluster system in a kind of unit, include distribution of services module,
Polylith network interface card, CPU and multiple hard disks, the CPU include the multiple of multiple calculating cores, the polylith network interface card and CPU
Multiple calculate nodes that core is corresponded in composition unit are calculated, the CPU is connected by data/address bus with multiple hard disks.
Further, the distribution of services module is DNS load balancing modules.
Further, the distribution of services module is HTTP scheduler modules.
Further, the plurality of hard disk is SSD hard disks.
Further, during many of the CPU calculates core, each calculates core with least one piece hard disk binding.
Another technical scheme of the present invention is:Computing cluster control method in a kind of unit, includes following step
Suddenly:
A, core and polylith network interface card connection composition multiple calculate nodes being calculated the CPU in unit, each calculate node includes more
There is a CPU to calculate core and one piece of network interface card, each calculate node at least binds one piece of hard disk;
B, distribution of services module obtain application request, and request content is sent to corresponding calculate node;
Network interface card in C, calculate node receives request content and is forwarded to corresponding CPU calculating core and processed, and then passes through
Data/address bus is written and read operation to the hard disk bound.
Further, in step A, the CPU of each calculate node calculates core and is provided with origin server mark, described
Origin server is identified for distinguishing application request source.
Further, step B is specially:Distribution of services module obtains application request, and by request content through loading
Send to corresponding calculate node after balanced or HTTP scheduling.
Further, step B specifically includes following sub-step:
B1, distribution of services module obtain application request;
Application request is originated to be sent to CPU by the network interface card of each calculate node and calculates core by B2, distribution of services module;
B3, CPU calculate core and application request source and origin server are represented contrast, and comparing result is calculated for identical CPU
Core is by network interface card feedback result to distribution of services module;
B4, distribution of services module send the request content of application request to corresponding network interface card according to above-mentioned steps B3 feedback.
Further, in step A, the CPU of each calculate node calculates core and is additionally provided with dispatching priority mark, institute
State step B4 to be specially:
If there is multiple CPU to calculate core by network interface card feedback result to distribution of services module, distribution of services module in step B3
The request content of application request is sent to corresponding network interface card according to above-mentioned dispatching priority mark.
The invention has the beneficial effects as follows:Multiple calculating cores, multiple Network card setups are many in unit by present system
Individual calculate node, and connect and bind with multiple hard disks, the reasonable distribution of application request is realized by distribution of services module, low cost
High IO performances and efficient PC cluster in unit is realized, and perfect can be adapted to multiple access networks.
The present invention another beneficial effect be:The inventive method is in unit by multiple calculating cores, multiple Network card setups
For multiple calculate nodes, and connect and bind with multiple hard disks, the reasonable distribution of application request is realized by distribution of services module, it is low
Cost realizes high IO performances and efficient PC cluster in unit, and perfect can be adapted to multiple access networks.
Description of the drawings
Fig. 1 is system structure schematic block diagram in unit in the present invention;
Fig. 2 is cluster control method flow chart of steps in unit in the present invention.
Specific embodiment
Below in conjunction with the accompanying drawings the specific embodiment of the present invention is described further:
With reference to Fig. 1, computing cluster system in a kind of unit, include distribution of services module,
Polylith network interface card, CPU and multiple hard disks, the CPU include the multiple of multiple calculating cores, the polylith network interface card and CPU
Multiple calculate nodes that core is corresponded in composition unit are calculated, the CPU is connected by data/address bus with multiple hard disks.
The CPU can be adopted:1st, multiple processors, each processor is used as a calculating core;2nd, polycaryon processor is made
For multiple calculating cores;3rd, other many calculating core technologies.
It is further used as preferred embodiment, the distribution of services module is DNS load balancing modules.
It is further used as preferred embodiment, the distribution of services module is HTTP scheduler modules.
It is further used as preferred embodiment, the plurality of hard disk is SSD hard disks.
It is further used as preferred embodiment, during many of the CPU calculates core, each calculates core with least one piece
Hard disk is bound.
With reference to Fig. 2, computing cluster control method in a kind of unit, wherein in unit, computing cluster system structure can refer to figure
1, include following steps:
A, core and polylith network interface card connection composition multiple calculate nodes being calculated the CPU in unit, each calculate node includes more
There is a CPU to calculate core and one piece of network interface card, each calculate node at least binds one piece of hard disk;
CPU calculating core amounts wherein in unit are expansible, and network interface card quantity accordingly increases, and hard disk quantity also correspondingly can increase
Plus.
B, distribution of services module obtain application request, and request content is sent to corresponding calculate node;
Network interface card in C, calculate node receives request content and is forwarded to corresponding CPU calculating core and processed, and then passes through
Data/address bus is written and read operation to the hard disk bound;
The CPU calculates core and can bind one or more hard disk, but one piece of hard disk only binds unique calculating core;Calculate core
Can be bound by simple mapping relations between the heart and hard disk, when the calculating core bound in hard disk is not utilized, can spirit
Adjustment binding relationship living, the calculating core more, frequently higher so as to be assigned to I/O operation amount.
It is further used as preferred embodiment, in step A, the CPU of each calculate node calculates core and be provided with
Source server is identified, and the origin server is identified for distinguishing application request source.
Each node in unit only processes corresponding Data Source, calculates core appointing in calculating process so as to reduce
Business handoff procedure, improves disposal ability and I/O rate.
It is further used as preferred embodiment, step B is specially:Distribution of services module obtains application request, and
Request content is sent to corresponding calculate node after load balancing or HTTP scheduling.
For example with DNS load-balancing techniques, the two-wire that computing cluster system deployment as shown in Figure 1 is arrived is modal double
In line machine room, calculate node A(Including network interface card A and calculating core A)The network of telecommunications is connected by the network interface card A of oneself, is calculated
Node B connects the network of UNICOM accordingly by network interface card B, by the scheduling of DNS so that the user of telecommunications accesses calculate node
A, user's accessed node of UNICOM calculate B, whole computer can be caused to reach the utilization of resources of maximum, including network I/O, magnetic
Disk IO and CPU etc..
It is further used as preferred embodiment, step B specifically includes following sub-step:
B1, distribution of services module obtain application request;
Application request is originated to be sent to CPU by the network interface card of each calculate node and calculates core by B2, distribution of services module;
B3, CPU calculate core and application request source and origin server are represented contrast, and comparing result is calculated for identical CPU
Core is by network interface card feedback result to distribution of services module;
B4, distribution of services module send the request content of application request to corresponding network interface card according to above-mentioned steps B3 feedback.
It is further used as preferred embodiment, in step A, the CPU of each calculate node calculates core and is additionally provided with
Dispatching priority is identified, and step B4 is specially:
If there is multiple CPU to calculate core by network interface card feedback result to distribution of services module, distribution of services module in step B3
The request content of application request is sent to corresponding network interface card according to above-mentioned dispatching priority mark.The dispatching priority can root
Set according to access frequency.
It is more than that the preferable enforcement to the present invention is illustrated, but the invention is not limited to the enforcement
Example, those of ordinary skill in the art on the premise of without prejudice to spirit of the invention can also be made a variety of equivalents or be replaced
Change, the deformation or replacement of these equivalents are all contained in the application claim limited range.
Claims (10)
1. computing cluster system in a kind of unit, it is characterised in that:Include distribution of services module, polylith network interface card, CPU and multiple
Hard disk, the CPU include multiple calculating cores of multiple calculating cores, the polylith network interface card and CPU and correspond composition list
Multiple calculate nodes in machine, the CPU are connected by data/address bus with multiple hard disks.
2. computing cluster system in a kind of unit according to claim 1, it is characterised in that:The distribution of services module is
DNS load balancing modules.
3. computing cluster system in a kind of unit according to claim 1, it is characterised in that:The distribution of services module is
HTTP scheduler modules.
4. computing cluster system in a kind of unit according to claim 1, it is characterised in that:The plurality of hard disk is SSD
Hard disk.
5. computing cluster system in a kind of unit according to claim 1, it is characterised in that:Many of the CPU calculates core
Each calculates core with least one piece hard disk binding in the heart.
6. computing cluster control method in a kind of unit, it is characterised in that include following steps:
A, core and polylith network interface card connection composition multiple calculate nodes being calculated the CPU in unit, each calculate node includes more
There is a CPU to calculate core and one piece of network interface card, each calculate node at least binds one piece of hard disk;
B, distribution of services module obtain application request, and request content is sent to corresponding calculate node;
Network interface card in C, calculate node receives request content and is forwarded to corresponding CPU calculating core and processed, and then passes through
Data/address bus is written and read operation to the hard disk bound.
7. computing cluster control method in a kind of unit according to claim 6, it is characterised in that:It is every in step A
The CPU of individual calculate node calculates core and is provided with origin server mark, and the origin server is identified please for distinguishing application
Ask source.
8. computing cluster control method in a kind of unit according to claim 7, it is characterised in that step B is concrete
For:Distribution of services module obtains application request, and request content is sent to corresponding after load balancing or HTTP scheduling
Calculate node.
9. computing cluster control method in a kind of unit according to claim 7, it is characterised in that step B is concrete
Including following sub-step:
B1, distribution of services module obtain application request;
Application request is originated to be sent to CPU by the network interface card of each calculate node and calculates core by B2, distribution of services module;
B3, CPU calculate core and application request source and origin server are represented contrast, and comparing result is calculated for identical CPU
Core is by network interface card feedback result to distribution of services module;
B4, distribution of services module send the request content of application request to corresponding network interface card according to above-mentioned steps B3 feedback.
10. computing cluster control method in a kind of unit according to claim 9, it is characterised in that:It is every in step A
The CPU of individual calculate node calculates core and is additionally provided with dispatching priority mark, and step B4 is specially:
If there is multiple CPU to calculate core by network interface card feedback result to distribution of services module, distribution of services module in step B3
The request content of application request is sent to corresponding network interface card according to above-mentioned dispatching priority mark.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610883006.7A CN106528488B (en) | 2016-10-09 | 2016-10-09 | Computing cluster system and control method in a kind of single machine |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610883006.7A CN106528488B (en) | 2016-10-09 | 2016-10-09 | Computing cluster system and control method in a kind of single machine |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106528488A true CN106528488A (en) | 2017-03-22 |
CN106528488B CN106528488B (en) | 2019-07-16 |
Family
ID=58331243
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610883006.7A Active CN106528488B (en) | 2016-10-09 | 2016-10-09 | Computing cluster system and control method in a kind of single machine |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106528488B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2024113925A1 (en) * | 2022-11-30 | 2024-06-06 | 苏州元脑智能科技有限公司 | Storage optimization method and system, device, and readable storage medium |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101631139A (en) * | 2009-05-19 | 2010-01-20 | 华耀环宇科技(北京)有限公司 | Load balancing software architecture based on multi-core platform and method therefor |
CN102999571A (en) * | 2012-11-09 | 2013-03-27 | 深圳市宜搜科技发展有限公司 | Realizing method for multiple nodes of single computer in cluster |
CN103634344A (en) * | 2012-08-23 | 2014-03-12 | 阿里巴巴集团控股有限公司 | Method and apparatus for unit operation multiple MySQL database examples |
-
2016
- 2016-10-09 CN CN201610883006.7A patent/CN106528488B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101631139A (en) * | 2009-05-19 | 2010-01-20 | 华耀环宇科技(北京)有限公司 | Load balancing software architecture based on multi-core platform and method therefor |
CN103634344A (en) * | 2012-08-23 | 2014-03-12 | 阿里巴巴集团控股有限公司 | Method and apparatus for unit operation multiple MySQL database examples |
CN102999571A (en) * | 2012-11-09 | 2013-03-27 | 深圳市宜搜科技发展有限公司 | Realizing method for multiple nodes of single computer in cluster |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2024113925A1 (en) * | 2022-11-30 | 2024-06-06 | 苏州元脑智能科技有限公司 | Storage optimization method and system, device, and readable storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN106528488B (en) | 2019-07-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US9882985B1 (en) | Data storage path optimization for internet of things computing system | |
US20190324819A1 (en) | Distributed-system task assignment method and apparatus | |
JP6290462B2 (en) | Coordinated admission control for network accessible block storage | |
WO2021126295A1 (en) | Request throttling in distributed storage systems | |
WO2015096681A1 (en) | Method and device for processing service access | |
CN104636077B (en) | Network control techology storage system and method for virtual machine | |
US9998531B2 (en) | Computer-based, balanced provisioning and optimization of data transfer resources for products and services | |
CN110602156A (en) | Load balancing scheduling method and device | |
CN109218355A (en) | Load equalizing engine, client, distributed computing system and load-balancing method | |
CN108900626B (en) | Data storage method, device and system in cloud environment | |
CN105007337A (en) | Cluster system load balancing method and system thereof | |
JP2013525931A (en) | Dynamic binding used for content delivery | |
US11307802B2 (en) | NVMe queue management multi-tier storage systems | |
WO2024016596A1 (en) | Container cluster scheduling method and apparatus, device, and storage medium | |
CN103297499A (en) | Scheduling method and system based on cloud platform | |
CN112600761B (en) | Resource allocation method, device and storage medium | |
CN102932271A (en) | Method and device for realizing load balancing | |
TW202112113A (en) | Resource sharing method, device, computer device and storage media | |
CN112363827A (en) | Multi-resource index Kubernetes scheduling method based on delay factors | |
US11709707B2 (en) | Low latency distributed counters for quotas | |
CN104809026B (en) | A kind of method that CPU computing resources are borrowed using remote node | |
Shu-Jun et al. | Optimization and research of hadoop platform based on fifo scheduler | |
CN104750614B (en) | Method and apparatus for managing memory | |
CN106528488A (en) | Computation cluster system in single machine and control method | |
CN110069319A (en) | A kind of multiple target dispatching method of virtual machine and system towards cloudlet resource management |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |