CN105338433A - Method for parallelly calculating broadcast communication - Google Patents
Method for parallelly calculating broadcast communication Download PDFInfo
- Publication number
- CN105338433A CN105338433A CN201510671311.5A CN201510671311A CN105338433A CN 105338433 A CN105338433 A CN 105338433A CN 201510671311 A CN201510671311 A CN 201510671311A CN 105338433 A CN105338433 A CN 105338433A
- Authority
- CN
- China
- Prior art keywords
- rack
- broadcast
- data
- cabinet
- free
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04Q—SELECTING
- H04Q11/00—Selecting arrangements for multiplex systems
- H04Q11/0001—Selecting arrangements for multiplex systems using optical switching
- H04Q11/0003—Details
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B10/00—Transmission systems employing electromagnetic waves other than radio-waves, e.g. infrared, visible or ultraviolet light, or employing corpuscular radiation, e.g. quantum communication
- H04B10/11—Arrangements specific to free-space transmission, i.e. transmission through air or vacuum
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- Electromagnetism (AREA)
- Signal Processing (AREA)
- Optical Communication System (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a method for parallelly calculating broadcast communication. All nodes whose broadcasting data should be sent send data to line head switches. Free light emission devices are arranged on cabinet tops of the line head switches to convert signals in fibers into free space light. Free light dividing and receiving devices are arranged on the cabinets for receiving data to divide and receive light so that the free space light is sent to all receiving ends on physical layers via the free light dividing and receiving devices via broadcast communication among the cabinets. Broadcast communication among inner servers of the cabinets is transported via wired links. The method is advantageous in that broadcast time delay can be effectively reduced and parallel calculating performance can be improved.
Description
Technical field
The present invention is mainly concerned with the field of broadcast communication in massive parallel processing, refers in particular to a kind of method of parallel computation broadcast communication.
Background technology
Massive parallel processing needs high field in computing capabilitys such as scientific algorithm, weather forecast, medical novel drugs researchs, and the economic development and strategic capability lifting of country have vital importance and functions.High-performance computer constantly promotes computing capability by technological innovation, and typical framework carries out parallel computation by computing cluster is collaborative.
The parallel programming models such as MPI (MessagePassingInterface) realize internodal cooperated computing by task is carried out division.One-to-all and all-to-one needs message distribution to converge to all nodes of the whole network in the operations such as wherein common broadcast, fence are synchronous.High-performance calculation is extremely responsive to the time delay of Message Transmission, and the deadline of reducing broadcast effectively can promote computational speed.Typical IBM BlueGene series super computers adopt independently tree network for accelerating broadcast communication.
Summary of the invention
The technical problem to be solved in the present invention is just: the technical problem existed for prior art, the invention provides a kind of method that effectively can reduce broadcast time delay, improve the parallel computation broadcast communication of Parallel Computing Performance
For solving the problems of the technologies described above, the present invention by the following technical solutions:
A method for parallel computation broadcast communication, data are sent to row head switch by all nodes sending broadcast data that needs, and by placing free light-emitting device on the rack top of row head switch, the signal in optical fiber are converted into free space optical; The rack receiving data places free light light splitting and receiving equipment carries out light splitting and reception, makes the broadcast communication between rack by free light light splitting and receiving equipment, free space optical can be sent to all receiving terminals in physical layer; Broadcast communication between interior of equipment cabinet server is then transmitted by wire link.
As a further improvement on the present invention: described free light-emitting device is optical fiber collimator.
As a further improvement on the present invention: described optical fiber collimator is one-way optical fiber collimater, the free space optical in a direction is used for being converted into through the signal that optical module transmitting terminal and optical fiber transmit.
As a further improvement on the present invention: described optical fiber collimator is bidirectional optical fiber collimater, the free space optical of both direction is used for being converted into through the signal that optical module transmitting terminal, optical fiber and optical fiber splitter transmit.
As a further improvement on the present invention: when described method completes the whole network broadcast by four switches, suppose that network size is that m arranges rack, often row has n rack, and concrete steps are as follows:
S100:(0,0) free light unidirectional emission equipment is placed in the horizontal direction respectively in rack top with vertical direction, send broadcast data for the rack to same row and same row; (on 0, j) 1≤j≤n rack top, horizontal direction places free light light splitting and receiving equipment for receiving data, vertical direction is placed free light unidirectional emission equipment, for being forwarded at same row by broadcast data; Free light light splitting and receiving equipment are placed in (i, j) 1≤i≤m, 1≤j≤n rack top, for receiving the broadcast data that vertical direction sent;
S200: data are sent to cabinet topcross by wired mode and change planes (i, j) by broadcast source node;
S300: cabinet topcross is changed planes and by wired mode, data is sent to the switch (i, 0) of same arrangement head rack;
S400: the cabinet topcross of rack (i, 0) cabinet topcross data being sent to (0,0) rack by wired mode of changing planes is changed planes;
S500:(0,0) the cabinet topcross of rack is changed planes the rack utilizing free light to be sent to by broadcast data same row all, and each rack, by free light light splitting and receiving equipment, receives the broadcast data that (0,0) rack sends;
S600:(0, j) 0≤j≤n rack forwards broadcast datas to all racks of same row, completes the whole network broadcast.
As a further improvement on the present invention: described method completes the whole network broadcast by three switches, suppose that network size is that m arranges rack, often row has n rack, and concrete steps are as follows:
S100:(i, 0) 1≤i≤(i of m rack on cabinet top, i) the two-way transmitter of free light is placed below individual position, place free light unidirectional emission equipment above, (the k on cabinet top, k) 0≤k≤m, k ≠ i places a position free light light splitting and receiving equipment; (i, j) 0≤i≤m, (i on 1≤j≤n cabinet top, i) the two-way transmitter of free light is placed below individual position, place free light light splitting and receiving equipment above, (k, the k) 0≤k≤m on cabinet top, k ≠ i place a position free light light splitting and receiving equipment;
S200: data are sent to cabinet topcross by wired mode and change planes (i, j) by broadcast source node;
S300: cabinet topcross is changed planes and by wired mode, data is sent to the switch (i, 0) of same arrangement head rack;
S400:(i, 0) the cabinet topcross of rack is changed planes the rack utilizing free light to be sent to by broadcast data same row all, and each rack, by free light light splitting and receiving equipment, receives the broadcast data that (i, 0) rack sends;
S500:(i, j) 0≤j≤n rack forwards broadcast datas to all racks of same row, completes the whole network broadcast.
Compared with prior art, the invention has the advantages that:
1, the method for parallel computation broadcast communication of the present invention, adopts free light to broadcast, and free light is compared with optical fiber, and free mpsi layer has natural broadcast characteristic, can carry broadcast communication application.Transmission has the feature of low delay, and the method utilizing free mpsi layer to carry out data broadcast can effectively reduce broadcast time delay, improves the performance of parallel computation.
2, the method for parallel computation broadcast communication of the present invention, utilizes the radio network of free light and other unicast traffics to isolate, and reduces network traffic load.
3, the method for parallel computation broadcast communication of the present invention, effectively can reduce the complexity of the hardware capabilitys such as switch network arbitration, reduces the capacity requirement transmitted.
4, the method for parallel computation broadcast communication of the present invention, utilizes free mpsi layer to carry out broadcast communication, namely replaces wire link with wireless link, can reduce wiring complexity, make system maintenance simple.
Accompanying drawing explanation
Fig. 1 is the schematic diagram that the present invention builds broadcast system in embody rule example.
Fig. 2 is the schematic diagram that the present invention adopts free light unidirectional emission equipment in embody rule example.
Fig. 3 is the schematic diagram that the present invention adopts free light light splitting and receiving equipment in embody rule example.
Fig. 4 is the schematic diagram that the present invention adopts the two-way transmitter of free light in embody rule example.
Fig. 5 is the present invention realizes the whole network broadcasting method in embodiment 1 principle schematic through four switches.
Fig. 6 is the present invention realizes the whole network broadcasting method in example 2 principle schematic through three switches.
Embodiment
Below with reference to Figure of description and specific embodiment, the present invention is described in further details.
The method of parallel computation broadcast communication of the present invention, for optimizing the method for parallel computation broadcast communication based on free space mpsi layer; In the method for the invention, data are sent to row head switch by all nodes sending broadcast data that needs, and by placing optical fiber collimator 1 on the rack top of row head switch, the signal in optical fiber are converted into free space optical A; The rack receiving data places ripple beam splitter 2 and carries out light splitting, make the broadcast communication between rack by ripple beam splitter 2, free space optical A can be sent to all receiving terminals in physical layer; Broadcast communication between interior of equipment cabinet server is then transmitted by wire link.
Concrete steps are:
S1: build broadcast system; See Fig. 1, the rack top of row head switch is placed free light unidirectional emission equipment (as: optical fiber collimator 1), the rack receiving data places free light light splitting and receiving equipment (as: ripple beam splitter 2); Wherein, optical fiber collimator 1 can be used for being converted into through the signal that optical module transmitting terminal 3 and optical fiber transmit the free space optical A (as shown in Figure 2) in a direction, or is used for being converted into through the signal that optical module transmitting terminal 3, optical fiber and optical fiber splitter 4 transmits the free space optical A (as shown in Figure 3) of both direction.Free space optical A can be carried out light splitting (as shown in Figure 4) by ripple beam splitter 2, makes the broadcast communication between rack by ripple beam splitter, free space optical can be sent to all receiving terminals in physical layer, i.e. optical module receiving terminal 5.
S2: the broadcast system built by step S1 utilizes free space optical A to carry out broadcast communication.
Embodiment 1: in an embody rule example, see Fig. 5, be complete the whole network broadcast (suppose that network size is that m arranges rack, often row has n rack) by four switches, detailed step is as follows:
S100:(0,0) free light unidirectional emission equipment is placed in the horizontal direction respectively in rack top with vertical direction, send broadcast data for the rack to same row and same row.(on 0, j) 1≤j≤n rack top, horizontal direction places free light light splitting and receiving equipment for receiving data, vertical direction is placed free light unidirectional emission equipment, for being forwarded at same row by broadcast data.Free light light splitting and receiving equipment are placed in (i, j) 1≤i≤m, 1≤j≤n rack top, for receiving the broadcast data that vertical direction sent.
S200: data are sent to cabinet topcross by wired mode and change planes (i, j) by broadcast source node;
S300: cabinet topcross is changed planes and by wired mode, data is sent to the switch (i, 0) of same arrangement head rack;
S400: the cabinet topcross of rack (i, 0) cabinet topcross data being sent to (0,0) rack by wired mode of changing planes is changed planes;
S500:(0,0) the cabinet topcross of rack is changed planes the rack utilizing free light to be sent to by broadcast data same row all, and each rack, by free light light splitting and receiving equipment, receives the broadcast data that (0,0) rack sends.
S600:(0, j) 0≤j≤n rack forwards broadcast datas to all racks of same row, completes the whole network broadcast.
Embodiment 2: in an embody rule example, see Fig. 6, be complete the whole network broadcast (suppose that network size is that m arranges rack, often row has n rack) by three switches, detailed step is as follows:
S100:(i, 0) 1≤i≤(i of m rack on cabinet top, i) the two-way transmitter of free light is placed below individual position, place free light unidirectional emission equipment above, (the k on cabinet top, k) 0≤k≤m, k ≠ i places a position free light light splitting and receiving equipment; (i, j) 0≤i≤m, (i on 1≤j≤n cabinet top, i) the two-way transmitter of free light is placed below individual position, place free light light splitting and receiving equipment above, (k, the k) 0≤k≤m on cabinet top, k ≠ i place a position free light light splitting and receiving equipment.
S200: data are sent to cabinet topcross by wired mode and change planes (i, j) by broadcast source node;
S300: cabinet topcross is changed planes and by wired mode, data is sent to the switch (i, 0) of same arrangement head rack;
S400:(i, 0) the cabinet topcross of rack is changed planes the rack utilizing free light to be sent to by broadcast data same row all, and each rack, by free light light splitting and receiving equipment, receives the broadcast data that (i, 0) rack sends.
S500:(i, j) 0≤j≤n rack forwards broadcast datas to all racks of same row, completes the whole network broadcast.
Below be only the preferred embodiment of the present invention, protection scope of the present invention be not only confined to above-described embodiment, all technical schemes belonged under thinking of the present invention all belong to protection scope of the present invention.It should be pointed out that for those skilled in the art, some improvements and modifications without departing from the principles of the present invention, should be considered as protection scope of the present invention.
Claims (6)
1. the method for a parallel computation broadcast communication, it is characterized in that, data are sent to row head switch by all nodes sending broadcast data that needs, and by placing free light-emitting device on the rack top of row head switch, the signal in optical fiber are converted into free space optical; The rack receiving data places free light light splitting and receiving equipment carries out light splitting and reception, makes the broadcast communication between rack by free light light splitting and receiving equipment, free space optical can be sent to all receiving terminals in physical layer; Broadcast communication between interior of equipment cabinet server is then transmitted by wire link.
2. the method for parallel computation broadcast communication according to claim 1, is characterized in that, described free light-emitting device is optical fiber collimator.
3. the method for parallel computation broadcast communication according to claim 2, is characterized in that, described optical fiber collimator is one-way optical fiber collimater, is used for being converted into through the signal that optical module transmitting terminal and optical fiber transmit the free space optical in a direction.
4. the method for parallel computation broadcast communication according to claim 2, it is characterized in that, described optical fiber collimator is bidirectional optical fiber collimater, is used for being converted into through the signal that optical module transmitting terminal, optical fiber and optical fiber splitter transmit the free space optical of both direction.
5. the method for parallel computation broadcast communication according to claim 1 and 2, is characterized in that, when described method completes the whole network broadcast by four switches, suppose that network size is that m arranges rack, often row has n rack, and concrete steps are as follows:
S100:(0,0) free light unidirectional emission equipment is placed in the horizontal direction respectively in rack top with vertical direction, send broadcast data for the rack to same row and same row; (on 0, j) 1≤j≤n rack top, horizontal direction places free light light splitting and receiving equipment for receiving data, vertical direction is placed free light unidirectional emission equipment, for being forwarded at same row by broadcast data; Free light light splitting and receiving equipment are placed in (i, j) 1≤i≤m, 1≤j≤n rack top, for receiving the broadcast data that vertical direction sent;
S200: data are sent to cabinet topcross by wired mode and change planes (i, j) by broadcast source node;
S300: cabinet topcross is changed planes and by wired mode, data is sent to the switch (i, 0) of same arrangement head rack;
S400: the cabinet topcross of rack (i, 0) cabinet topcross data being sent to (0,0) rack by wired mode of changing planes is changed planes;
S500:(0,0) the cabinet topcross of rack is changed planes the rack utilizing free light to be sent to by broadcast data same row all, and each rack, by free light light splitting and receiving equipment, receives the broadcast data that (0,0) rack sends;
S600:(0, j) 0≤j≤n rack forwards broadcast datas to all racks of same row, completes the whole network broadcast.
6. the method for parallel computation broadcast communication according to claim 1 and 2, is characterized in that, described method completes the whole network broadcast by three switches, and suppose that network size is that m arranges rack, often row has n rack, and concrete steps are as follows:
S100:(i, 0) 1≤i≤(i of m rack on cabinet top, i) the two-way transmitter of free light is placed below individual position, place free light unidirectional emission equipment above, (the k on cabinet top, k) 0≤k≤m, k ≠ i places a position free light light splitting and receiving equipment; (i, j) 0≤i≤m, (i on 1≤j≤n cabinet top, i) the two-way transmitter of free light is placed below individual position, place free light light splitting and receiving equipment above, (k, the k) 0≤k≤m on cabinet top, k ≠ i place a position free light light splitting and receiving equipment;
S200: data are sent to cabinet topcross by wired mode and change planes (i, j) by broadcast source node;
S300: cabinet topcross is changed planes and by wired mode, data is sent to the switch (i, 0) of same arrangement head rack;
S400:(i, 0) the cabinet topcross of rack is changed planes the rack utilizing free light to be sent to by broadcast data same row all, and each rack, by free light light splitting and receiving equipment, receives the broadcast data that (i, 0) rack sends;
S500:(i, j) 0≤j≤n rack forwards broadcast datas to all racks of same row, completes the whole network broadcast.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510671311.5A CN105338433B (en) | 2015-10-16 | 2015-10-16 | A kind of method of parallel computation broadcast communication |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510671311.5A CN105338433B (en) | 2015-10-16 | 2015-10-16 | A kind of method of parallel computation broadcast communication |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105338433A true CN105338433A (en) | 2016-02-17 |
CN105338433B CN105338433B (en) | 2019-01-08 |
Family
ID=55288653
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510671311.5A Active CN105338433B (en) | 2015-10-16 | 2015-10-16 | A kind of method of parallel computation broadcast communication |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105338433B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1298113A (en) * | 1999-11-30 | 2001-06-06 | 王宝根 | Infrared multispectral scanning thermal imager |
CN101487915A (en) * | 2008-01-17 | 2009-07-22 | 3M创新有限公司 | Optical apparatus holder |
CN103441942A (en) * | 2013-08-26 | 2013-12-11 | 重庆大学 | Data center network system and data communication method based on software definition |
CN204575907U (en) * | 2015-05-04 | 2015-08-19 | 深圳市鹏大光电技术有限公司 | A kind of forward direction optoelectric hybrid device of Erbium-Doped Fiber Amplifier (EDFA) |
CN104904143A (en) * | 2012-11-26 | 2015-09-09 | 维斯柯科技有限公司 | Methods and systems for passive optical switching |
CN104949770A (en) * | 2015-07-13 | 2015-09-30 | 天津津航技术物理研究所 | TDLAS (tunable diode laser absorption spectroscopy) gas temperature measurement and detection device |
-
2015
- 2015-10-16 CN CN201510671311.5A patent/CN105338433B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1298113A (en) * | 1999-11-30 | 2001-06-06 | 王宝根 | Infrared multispectral scanning thermal imager |
CN101487915A (en) * | 2008-01-17 | 2009-07-22 | 3M创新有限公司 | Optical apparatus holder |
CN104904143A (en) * | 2012-11-26 | 2015-09-09 | 维斯柯科技有限公司 | Methods and systems for passive optical switching |
CN103441942A (en) * | 2013-08-26 | 2013-12-11 | 重庆大学 | Data center network system and data communication method based on software definition |
CN204575907U (en) * | 2015-05-04 | 2015-08-19 | 深圳市鹏大光电技术有限公司 | A kind of forward direction optoelectric hybrid device of Erbium-Doped Fiber Amplifier (EDFA) |
CN104949770A (en) * | 2015-07-13 | 2015-09-30 | 天津津航技术物理研究所 | TDLAS (tunable diode laser absorption spectroscopy) gas temperature measurement and detection device |
Also Published As
Publication number | Publication date |
---|---|
CN105338433B (en) | 2019-01-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102546402B (en) | Method for supporting distributed key-based processes and computing entity | |
CN106688208A (en) | Network communications using pooled memory in rack-scale architecture | |
EP4258720A3 (en) | Small cell network architecture for servicing multiple network operations | |
CN103970214A (en) | Heterogeneous acceleration blade type computer system architecture | |
CN104170319A (en) | Management of distributed fabric system | |
CN104579696B (en) | A kind of optical fiber switch system and a kind of optical fiber switch | |
CN104704484A (en) | Communicating tuples in message | |
CN104580527B (en) | A kind of more I/O high density multi-node server system design methods of cloud service-oriented device application | |
CN103024071A (en) | Quantum stream media cloud computing based cloud aggregation community system | |
CN104954439B (en) | A kind of Cloud Server and its node interconnected method, cloud server system | |
CN103345458A (en) | Multi-FPGA interconnection structure and logic partitioning method oriented to high-performance computing | |
CN112242963B (en) | Rapid high-concurrency neural pulse data packet distribution and transmission method and system | |
CN105338433A (en) | Method for parallelly calculating broadcast communication | |
CN201387555Y (en) | Comprehensive remote control monitoring system | |
CN103929365B (en) | A kind of SiteServer LBS and method suitable for UDP service | |
CN101217564B (en) | A parallel communication system and the corresponding realization method of simple object access protocol | |
US10805233B2 (en) | Fractal-tree communication structure and method, control apparatus and intelligent chip | |
CN103957473A (en) | Data center server interconnection system, building and addressing method and cabinets | |
CN105359432B (en) | The communication system and communication means of a kind of data center | |
CN110635931B (en) | Link group configuration method and device | |
CN108934029A (en) | The acceleration distributed optimization algorithm rebuild towards perception big data | |
US7751344B2 (en) | Computer system and method using a kautz-like digraph to interconnect computer nodes and having control back channel between nodes | |
CN102737010B (en) | Parallel matrix multiplication method and system with Mohr diagram serving as topological structure | |
CN108293006A (en) | Technology for tracking disordered networks grouping | |
CN105577769A (en) | Resource distribution system in multipath high-end computer system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |