CN105959239B - The communication means of three-dimensional network on mating plate - Google Patents
The communication means of three-dimensional network on mating plate Download PDFInfo
- Publication number
- CN105959239B CN105959239B CN201610255814.9A CN201610255814A CN105959239B CN 105959239 B CN105959239 B CN 105959239B CN 201610255814 A CN201610255814 A CN 201610255814A CN 105959239 B CN105959239 B CN 105959239B
- Authority
- CN
- China
- Prior art keywords
- node
- resource reservation
- grouping
- destination node
- network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/70—Admission control; Resource allocation
- H04L47/72—Admission control; Resource allocation using reservation actions during connection setup
- H04L47/724—Admission control; Resource allocation using reservation actions during connection setup at intermediate nodes, e.g. resource reservation protocol [RSVP]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L49/00—Packet switching elements
- H04L49/10—Packet switching elements characterised by the switching fabric construction
- H04L49/109—Integrated on microchip, e.g. switch-on-chip
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a kind of communication means of three-dimensional network on mating plate, mainly solve the problems, such as that communication link utilization rate is low, other nodes of single node occlusive effects communicate, communication delay is big low with network throughput in existing three-dimensional network on mating plate.Its technical solution is: by taking three-dimensional Mesh and Torus structure as an example, when resource reservation, which is grouped in intermediate node, to be blocked, resource reservation grouping is retransmitted according to network real-time condition, and the condition of resource reservation packet retransmissions present is optimized during carrying out resource reservation packet retransmissions present, the present invention overcomes communication blocking in three-dimensional Mesh and three-dimensional Torus structure is larger, the problem of other nodes of the occlusive effects of single node communicate, relieve the deadlock that three-dimensional Torus structure is generated using most short dimension sequence routing algorithm, communication link utilization rate and handling capacity are improved in the case where energy consumption increases less, reduce delay, it can be used for realizing the efficient communication on mating plate between IP kernel.
Description
Technical field
The invention belongs to field of communication technology, in particular to a kind of communication means of network-on-chip can be used for realizing mating plate
Efficient communication between upper IP kernel.
Background technique
During system on chip develops towards multinucleation and isomerization direction, the concept of network-on-chip is suggested,
Its core concept is fused to computer networking technology in chip design, architecturally comes thoroughly to improve bus structures band
The a series of problems come.But as the IP kernel number on monolithic integrated optical circuit reaches many-core even thousand cores, integrated circuit dimensions enter
The problems such as after nanometer era, clock frequency is higher and higher, the parasitic capacitance of electric interconnection technique, signal cross-talk will become very aobvious
It writes.Light network technology is due to high bandwidth, low-power consumption, low delay, and the advantages such as small crosstalk and electromagnetic interference can efficiently solve
A series of bottleneck problems in network-on-chip tradition electrical interconnection technology field.Therefore, light network technology has become network-on-chip times
A kind of concerned technology.
As nucleus number further increases, chip size is further reduced, in the case where process reduction is limited, two dimension
Network-on-chip integration density, communication bandwidth, scalability and in terms of highlight bottleneck, therefore three-dimensional light piece is surfed the Internet
Network comes into being, and solves above-mentioned problems by the 3 D stereo interconnected communication of three-dimensional network on mating plate.
Currently, lacking low cost, efficient Optical buffer device on piece, while light logic processing technique is still immature
Under the conditions of, traditional optical circuit exchange is since its data transmission bauds is fast, it is reliable not need on piece Optical buffer, information packet transmissions
The advantages that be widely used in network on mating plate.In optical circuit exchange, the communication between source node and destination node needs only
Enjoy link circuit resource.It has disadvantages such that the utilization rate of link is lower, with the increase of network load, network congestion can be relatively tighter
Weight.
Two kinds of network architectures of three-dimensional Mesh and Torus have the characteristics that topology rule, structure are simple, are easily achieved, and are mesh
The preceding widely used framework of network on three-dimensional chip.When optical circuit exchange is applied to three-dimensional karnaugh map framework, link setup is grouped in
Once discovery output port is occupied during reserving light path, present node can only be blocked in and wait port release.Meanwhile
Since the last period path of the link setup packet reservation does not discharge, so that this section of resource oneself can neither use, and cannot discharge
It is used to other communication nodes.This greatly reduce the utilization rate of link and cause the obstruction of single node diffuse to network its
His node.When optical circuit exchange is applied to the three-dimensional Torus network architecture, other than having problem above, in communication process also
It is easy to produce deadlock, seriously reduces communication efficiency.To mitigate problem above bring harm, Gu Huaxi et al. " is realizing mating plate
The low-resistance plug communications router and communication means of upper network ", in the patent of Patent No. ZL201110330001.9, for two dimension
Mesh network proposes the communication means of data collection grouping: when link setup, which is grouped in intermediate node, to be blocked, data temporarily being returned
Intermediate node is received, destination node is then transmitted to by intermediate node again.Although the method overcomes single to a certain extent
The obstruction of node may influence the problem of other nodes communication, but since on piece lacks inexpensive, efficient Optical buffer device,
Light logic processing technique is still immature simultaneously, and data of every recycling just need to be transformed into electric signal by an optical signal, electricity
Signal is in the process for being converted into optical signal, this meeting is so that energy consumption is very big.If the method is used for 3-D mesh network structure,
The above problem can be protruded more, and the method cannot be used for solving Deadlock in Torus network structure.
Summary of the invention
The purpose of the present invention is being directed to the deficiency of above-mentioned prior art, a kind of communication party of three-dimensional network on mating plate is proposed
Method solves the problems, such as that other nodes of single node occlusive effects communicate, improves resource utilization to release the deadlock in communication,
Reduce delay, improve network throughput.
To achieve the above object, technical solution of the present invention includes the following:
(1) netinit
1a) parameter initialization: the turn around time weight β of setting obstruction confirmation grouping hinders other resource reservation groupings
Number passback threshold value NmaxAnd source node IP kernel receives the waiting for retransmitting resource reservation grouping after obstruction confirmation is grouped again
Time Tip；
1b) construct network: using the electric control network and optical transport network of router building three-dimensional light on piece structure；
It 1c) establishes coordinate system: establishing the three-dimensional system of coordinate XYZ of electric control network and optical transport network, sat according to the three-dimensional
Mark is the source node location (x for determining present communicationss, ys, zs) and destination node position (xd, yd, zd)；
(2) source node generates resource reservation grouping, and source node (x is marked in the groupings, ys, zs), destination node (xd,
yd, zd) location information and resource reservation grouping serial number, and the resource reservation grouping of these information will be carried by electric control net
Network is transmitted to destination node, and reserves optical link resource on the way during transmission；
(3) in the intermediate node of electric control network according to the output port M of routing algorithm computing resource reservation packet, inquiry
Whether output port M is locked: if locked, inquiring the resource reservation grouping for locking the port, this resource reservation is grouped
The number for blocking other groupings adds 1, executes step (4)；Otherwise, output port M is locked with intermediate node, and records resource reservation
The input/output port information of grouping and the locking moment T of output port Mlock, forward the resource reservation to be grouped, execute step
(5)；
(4) resource reservation packet retransmissions present process is carried out
4a) according to present node position (xc, yc, zc) and resource reservation grouping carry source node location (xs, ys, zs),
Calculate the hop count H between present node and source nodecs；According to HcsJudge whether present node is source node: if Hcs=0, then when
Front nodal point is source node, executes step 4i)；Otherwise, step 4b is executed)；
4b) according to present node position (xc, yc, zc) and resource reservation grouping carry destination node position (xd, yd,
zd), calculate the hop count H between present node and point of destinationcd；According to HcdJudge that next-hop is destination node: if Hcd≤ 1,
Then present node is destination node, executes step 4i)；Otherwise, step 4c is executed)；
4c) query resource reservation packet has blocked the number N of other groupingbIf Nb>Nmax, then follow the steps 4g)；If 1
≤Nb≤Nmax, then follow the steps 4d)；If Nb=0, then follow the steps 4i)；
4d) according to the hop count H between present node and destination nodecdIn output port M inquiry present node record
Port average lock delay Tnew_avg；If inquiry is recorded as sky, which prolongs Tnew_avg=0；It calculates to wait and be somebody's turn to do
Be delayed T needed for the release of portwait:
Twait=Tnew_avg-(Tcur-Tlock)
In formula, TcurFor current time, TlockAt the time of when locked for the output port, Tnew_avgIndicate that the port is being worked as
Front nodal point is apart from destination node HcdThe average delay of required locking when jump executes step 4e)；
4e) according to the hop count H between present node and source nodecsCalculate the time cost of retransmission resource reservation packet
Tprice:
Tprice=Ttear_mod+Tip+Tset_mod+Hcs×Tprocess
In formula, Ttear_modFor the transmission delay of obstruction confirmation grouping, TipFor source node receive obstruction confirmation grouping after and again
The new interval delay for generating resource reservation grouping, Tset_modFor the transmission delay of resource reservation grouping, TprocessFor resource reservation point
Group executes step 4f by delay needed for each intermediate node in electric control network)；
4f) comparison step 4d) obtain delay TwaitThe time cost T obtained with step 4e)priceIf: Twait> Tprice,
Then meet retransmission condition, execute step 4g), otherwise, execute step 4i)；
4g) present node generates obstruction confirmation grouping, and obstruction confirmation grouping, which carries, tears chain information, source node and purpose section open
The location information and resource reservation grouping serial number of point, are sent to source node by the optical link reserved, obstruction confirmation grouping
What is carried tears chain information open for along the resource of road release resource reservation packet reservation, executes step 4h)；
After 4h) source node receives obstruction confirmation grouping, in T lateripMoment regenerates resource reservation grouping, the resource
Reservation packet carries the location information and resource reservation grouping serial number that obstruction confirmation is grouped the source node, destination node taken back, leads to
It crosses electric control network to transmit to destination node, and reserves path on the way during transmission, return step 3；
4i) the medium port M to be output of caching that resource reservation is grouped in present node is released；
(5) according to the hop count H between present node and point of destinationcdJudge whether resource reservation grouping reaches destination node,
If HcdIt is not zero, then resource reservation is grouped the node return step 3 that misses one's purpose；Otherwise, resource reservation is grouped the section that achieves the goal
Point executes step 6；
(6) after resource reservation grouping reaches destination node, respond packet is generated by destination node, which carries should
Resource reservation grouping serial number is sent to source node by the optical link reserved；
(7) source node checks the serial number that respond packet carries after receiving respond packet, and extracts this according to this serial number
Respective data packet in ground data packet queue transmits the data grouping to destination node by the optical link reserved；
(8) data grouping is transmitted, the tail portion of data grouping, which carries, tears chain information open, in the path along reservation by intermediate
When node, this, which tears chain information open, will notify optical router to discharge the resource being reserved, and it is flat that the port is updated when unlocking output port M
Locking delay Tnew_avg；Data grouping is transmitted to destination node along the optical link made an appointment in advance；
(9) it after destination node receives data grouping, is sent in local IP kernel and handles, complete entire communication process.
The present invention has the advantage that compared with current art
1. the present invention is grouped resource reservation according to network real time environment and carries out dynamic re-transmission, conventional three-dimensional mating plate is overcome
Block the problem of larger, single node obstruction may influence the communication of other nodes in upper system, solves some three-dimensional networks
Deadlock in communication reduces delay in the case where energy consumption increases less, improves handling capacity.
2. the present invention combine resource reservation be grouped needed for the waiting time, with destination node at a distance from and resource reservation be grouped
Block the factors such as the number of other resource reservations grouping, weight is grouped to resource reservation during carrying out resource reservation packet retransmissions present
The condition of biography is optimized, and overcomes existing communication method and ignores the actual conditions of communication, do not have the problem of flexibility.
Detailed description of the invention
Fig. 1 is implementation flow chart of the invention.
Specific embodiment
Embodiments of the present invention are described in further detail with reference to the accompanying drawing.
The present invention solves three-dimensional Mesh and three-dimensional Torus structure uses light by taking three-dimensional Mesh and three-dimensional Torus structure as an example
The problem of link utilization is lower when circuit switching, other nodes of single node occlusive effects communicate, and solve three-dimensional Torus knot
Using the Deadlock of optical circuit exchange in structure, the utilization rate of Internet resources is effectively improved, to reduce communication delay, improves net
Network handling capacity.
Referring to Fig.1, steps are as follows for realization of the invention:
Step 1: netinit.
The initialization of network includes parameter initialization, constructs network and establishes three steps of coordinate system:
1a) parameter initialization: resource reservation is grouped into the transmission reservation optical link resource of data grouping, when resource reservation point
When group is blocked in electric control network, it is desirable to by discharging the last period link of the resource reservation packet reservation in time, with
Just other data groupings can use, and reduce network congestion.Therefore need to select rule and the setting of resource reservation packet retransmissions present
For parameter to being blocked whether resource reservation grouping executes re-transmission and determine, design parameter includes obstruction confirmation grouping turn around time
Weight β hinders other resource reservation grouping number threshold value NmaxAnd source node IP kernel is received after obstruction confirmation is grouped and is weighed again
Send out waiting time T needed for the resource reservation is groupedip；
It is 0≤β≤1, the turn around time weight and the delay T for waiting port release that β, which meets condition,waitCorrelation, NmaxAccording to
Network average number of hops determines its specific value, TipIts specific value is determined according to current Network uninterrupted；
1b) construct network: using the electric control network and light of router building three-dimensional Mesh structure, three-dimensional Torus structure
Transmission network, the router have 7 pairs of input/output ports, and wherein local input output port passes through network interface and IP kernel phase
Even, input/output port eastwards, westwards, northwards, southwards, up and down is connected with adjacent router respectively；
It 1c) establishes coordinate system: in electric control network, using the lower left corner as origin, being established according to right-hand rule three-dimensional
Coordinate system XYZ, node coordinate value are (x, y, z)；Source node location (the x of present communications is determined according to the three-dimensional system of coordinates, ys,
zs) and destination node position (xd, yd, zd)；The coordinate system and electric control network of optical transport network are overlapped.
Step 2: generating resource reservation grouping.
Source node generates resource reservation grouping, and source node (x is marked in the groupings, ys, zs), destination node (xd, yd,
zd) location information and resource reservation grouping serial number；
Location information and resource reservation grouping serial number will be carried and transmitted by electric control network to destination node, and transmitted
During reserve optical link resource on the way.
Step 3: transfer resource reservation packet.
Transfer resource reservation packet includes calculating whether locked two steps of output port M and inquiry output port M:
3a) calculate output port M:
In different structures, when resource reservation grouping is transferred to intermediate node, according to different structures using different
Routing algorithm calculates its required output port, it may be assumed that
Output port M is calculated in three-dimensional Mesh structure, is the intermediate node by electric control network using XYZ dimension sequence routing
The output port M of algorithm computing resource reservation packet；
In three-dimensional Torus structure, the intermediate node of electric control network is pre- using dimension sequence shortest route algorithm computing resource
The output port M being about grouped；
Wherein the XYZ dimension sequence routing algorithm is the three-dimensional common routing algorithm of Mesh structure, which has easy
It realizes, is simple, without livelock and the advantages of deadlock freedom.In the routing algorithm, grouping is tieed up along electric control network X carry out road first
By until reaching and destination node X-coordinate same node point position (xd, ys, zs), then the grouping is along electric control network Y Wei Lu
By until reaching node location (x identical with destination node Y coordinated, yd, zs), the finally grouping is along electric control network Z Wei Lu
By to destination node；
Wherein the dimension sequence shortest route is common routing algorithm in Torus network.In the routing algorithm, grouping is first
First the hop count minimum direction along electric control network X dimension from Y dimension where destination node is routed, until arrival and destination node
The identical node location of Y coordinate (xd, ys, zs), then the grouping is along electric control network Y dimension to Z where destination node
The dimension least direction of hop count is routed, until reaching node location (x identical with the Z coordinate of destination noded, yd, zs), most
The grouping is routed along electric control network Z dimension to the least direction of destination node hop count afterwards, until destination node；
3b) whether inquiry output port M is locked: if locked, the resource reservation grouping for locking the port is inquired, it will
The number that other groupings are blocked in the grouping of this resource reservation adds 1, executes step (4)；Otherwise, output port M is locked with intermediate node,
And record the input/output port information of resource reservation grouping and the locking moment T of output port Mlock, forward the resource reservation
Grouping executes step (5).
Step 4: carrying out resource reservation packet retransmissions present process.
4a) according to present node position (xc, yc, zc) and resource reservation grouping carry source node location (xs, ys, zs),
Calculate the hop count H between present node and source nodecs: Hcs=| xc-xs|+|yc-ys|+|zc-zs|；According to HcsProsthomere is worked as in judgement
Whether point is source node: if Hcs=0, then present node is source node, executes step 4i)；Otherwise, step 4b is executed)；
4b) according to present node position (xc, yc, zc) and resource reservation grouping carry destination node position (xd, yd,
zd), calculate the hop count H between current node and destination nodecd: Hcd=| xc-xd|+|yc-yd|+|zc-zd|；According to HcdJudgement
Next-hop is destination node: if Hcd≤ 1, then present node is destination node, executes step 4i)；Otherwise, step is executed
4c)；
4c) query resource reservation packet blocks the number N of other groupingbIf Nb> Nmax, thens follow the steps 4g)；If 1≤
Nb≤Nmax, then follow the steps 4d)；If Nb=0, then follow the steps 4i)；
4d) according to the hop count H between present node and destination nodecdIn output port M inquiry present node record
Port average lock delay Tnew_avg；If inquiry is recorded as sky, port average lock delay Tnew_avg=0；It calculates and waits
Be delayed T needed for the release of the portwait: Twait=Tnew_avg-(Tcur-Tlock),
In formula, TcurFor current time, TlockAt the time of when locked for the output port, Tnew_avgIndicate that the port is being worked as
Front nodal point is apart from destination node HcdThe average delay of required locking, T when jumpwaitValue is more than or equal to zero, if calculating to obtain TwaitIt is less than
Zero, then take Twait=0；
4e) according to the hop count H between present node and source nodecsCalculate the time cost of retransmission resource reservation packet
Tprice:
Tprice=Ttear_mod+Tip+Tset_mod+Hcs×Tprocess
In formula, Ttear_modFor the transmission delay of obstruction confirmation grouping, TipFor source node receive obstruction confirmation grouping after and again
The new interval delay for generating resource reservation grouping, Tset_modFor the transmission delay of resource reservation grouping, TprocessFor resource reservation point
Group is by delay needed for each intermediate node in electric control network
4f) comparison step 4d) obtain delay TwaitThe time cost T obtained with step 4e)priceIf: Twait> Tprice,
Then meet retransmission condition, execute step 4g), otherwise, execute step 4i)；
4g) present node generates obstruction confirmation grouping, and the obstruction is confirmed grouping hair by the optical link by having reserved
Toward source node, obstruction confirmation grouping, which carries, tears chain information, the location information of source node, the location information of destination node and money open
Source reservation packet serial number, the resource torn chain information open and will discharge resource reservation packet reservation along road that obstruction confirmation grouping carries,
Execute step 4h)；
After 4h) source node receives obstruction confirmation grouping, in T lateripMoment regenerates resource reservation grouping, the resource
Reservation packet carries source node location information, destination node location information and the resource reservation grouping that obstruction confirmation grouping is taken back
Serial number is transmitted to destination node by electric control network, and reserves path on the way during transmission, return step 3；
4i) resource reservation, which is grouped in the caching of present node, waits port M to be released.
Step 5: judging whether resource reservation grouping reaches destination node.
According to the hop count H between present node and point of destinationcdJudge whether resource reservation grouping reaches destination node, if
HcdIt is not zero, then resource reservation is grouped the node return step 3 that misses one's purpose；Otherwise, resource reservation is grouped the node that achieves the goal,
Execute step 6.
Step 6: generating respond packet.
After resource reservation grouping reaches destination node, respond packet is generated by destination node, which carries the money
Source reservation packet serial number is sent to source node by the optical link reserved.
Step 7: sending data grouping.
Source node checks the serial number that respond packet carries after receiving respond packet, and is extracted locally according to this serial number
Respective data packet in data packet queue sends the data grouping to the optical link reserved.
Step 8: transmission data grouping.
The tail portion of data grouping, which carries, tears chain information open, and when passing through intermediate node along the path of reservation, this tears chain letter open
Breath will notify optical router to discharge the resource being reserved, and port average lock delay is updated when unlocking output port M
Tnew_avg
Tnew_avg=(1- β) (Tunlock-Tlock)+βTold_avg
Wherein, TunlockAt the time of indicating unlock output port M, TlockAt the time of locked for output port M, (Tunlock-
Tlock) indicate that this has reserved HcdJump resource reservation grouping from request the router this port to release port delay,
Told_avgIt indicates to have been reserved H before output port McdThe average delay of the resource reservation grouping locking of jump, its initial value are
0, β indicates Told_avgWith (Tunlock-Tlock) weight, value be 0 to 1 between；Data grouping is passed along the optical link made an appointment in advance
Transport to destination node.
Step 9: receiving data grouping
After destination node receives data grouping, it is sent in local IP kernel and handles, complete entire communication process.
Above description is only example of the present invention, does not constitute any limitation of the invention.Obviously for this
It, all may be without departing substantially from the principle of the invention, structure after having understood the content of present invention and principle for the professional in field
In the case of, carry out various modifications and change in form and details, such as the present invention can be used for except three-dimensional Mesh and Tours with
Outer network, but these modifications and changes based on inventive concept are still within the scope of the claims of the present invention.
Claims (4)
1. the communication means of three-dimensional network on mating plate, comprising:
(1) netinit
1a) parameter initialization: the turn around time weight β of setting obstruction confirmation grouping hinders other resource reservation grouping numbers to return
Pass threshold value NmaxAnd source node IP kernel receives the waiting time for retransmitting resource reservation grouping after obstruction confirmation is grouped again
Tip；
1b) construct network: using the electric control network and optical transport network of router building three-dimensional light on piece structure；
It 1c) establishes coordinate system: the three-dimensional system of coordinate XYZ of electric control network and optical transport network is established, according to the three-dimensional system of coordinate
Determine the source node location (x of present communicationss, ys, zs) and destination node position (xd, yd, zd)；
(2) source node generates resource reservation grouping, and source node (x is marked in the groupings, ys, zs), destination node (xd, yd, zd)
Location information and resource reservation grouping serial number, and the resource reservation groupings of these information will be carried by electric control network to mesh
Node-node transmission, and reserve optical link resource on the way during transmission；
(3) in the intermediate node of electric control network according to the output port M of routing algorithm computing resource reservation packet, inquiry output
Whether port M is locked: if locked, inquiring the resource reservation grouping for locking the port, this resource reservation is grouped and is blocked
The number of other groupings adds 1, executes step (4)；Otherwise, output port M is locked with intermediate node, and records resource reservation grouping
Input/output port information and output port M locking moment Tlock, forward the resource reservation to be grouped, execute step (5)；
(4) resource reservation packet retransmissions present process is carried out
4a) according to present node position (xc, yc, zc) and resource reservation grouping carry source node location (xs, ys, zs), it calculates
Hop count H between present node and source nodecs；According to HcsJudge whether present node is source node: if Hcs=0, then work as prosthomere
Point is source node, executes step 4i)；Otherwise, step 4b is executed)；
4b) according to present node position (xc, yc, zc) and resource reservation grouping carry destination node position (xd, yd, zd), meter
Calculate the hop count H between current node and destination nodecd；According to HcdJudge that next-hop is destination node: if Hcd≤ 1, then when
Front nodal point is destination node, executes step 4i)；Otherwise, step 4c is executed)；
4c) query resource reservation packet has blocked the number N of other groupingbIf Nb>Nmax, then follow the steps 4g)；If 1≤Nb
≤Nmax, then follow the steps 4d)；If Nb=0, then follow the steps 4i)；
4d) according to the hop count H between present node and destination nodecdWith the end in output port M inquiry present node record
Mouth average lock delay Tnew_avg；If inquiry is recorded as sky, port average lock delay Tnew_avg=0；It calculates and waits the end
Be delayed T needed for mouth releasewait:
Twait=Tnew_avg-(Tcur-Tlock)
In formula, TcurFor current time, TlockAt the time of when locked for the output port, Tnew_avgIndicate that prosthomere is being worked as in the port
Point is apart from destination node HcdThe average delay of required locking when jump executes step 4e)；
4e) according to the hop count H between present node and source nodecsCalculate the time cost T of retransmission resource reservation packetprice:
Tprice=Ttear_mod+Tip+Tset_mod+Hcs×Tprocess
In formula, Ttear_modFor the transmission delay of obstruction confirmation grouping, TipFor source node receive obstruction confirmation grouping after with produce again
The interval of production-goods source reservation packet is delayed, Tset_modFor the transmission delay of resource reservation grouping, TprocessIt is grouped and passes through for resource reservation
Delay needed for each intermediate node in electric control network is crossed, step 4f is executed)；
4f) comparison step 4d) obtain delay TwaitThe time cost T obtained with step 4e)priceIf: Twait> Tprice, then full
Sufficient retransmission condition executes step 4g), otherwise, execute step 4i)；
4g) present node generates obstruction confirmation grouping, and obstruction confirmation grouping, which carries, tears chain information, source node and destination node open
Location information and resource reservation grouping serial number are sent to source node by the optical link reserved, and obstruction confirmation grouping carries
Tear open chain information will along road discharge resource reservation packet reservation resource, execute step 4h)；
After 4h) source node receives obstruction confirmation grouping, in T lateripMoment regenerates resource reservation grouping, the resource reservation
Grouping carries the location information and resource reservation grouping serial number that obstruction confirmation is grouped the source node, destination node taken back, passes through electricity
It controls network to transmit to destination node, and reserves path on the way during transmission, return step (3)；
4i) the medium port M to be output of caching that resource reservation is grouped in present node is released；
(5) according to the hop count H between present node and point of destinationcdJudge whether resource reservation grouping reaches destination node, if Hcd
It is not zero, then resource reservation is grouped the node return step 3 that misses one's purpose；Otherwise, resource reservation is grouped the node that achieves the goal, and holds
Row step (6)；
(6) after resource reservation grouping reaches destination node, respond packet is generated by destination node, which carries the resource
Reservation packet serial number is sent to source node by the optical link reserved；
(7) source node checks the serial number that respond packet carries after receiving respond packet, and extracts local number according to this serial number
According to the respective data packet in packet queue, which is transmitted to destination node by the optical link reserved；
(8) data grouping is transmitted, the tail portion of data grouping, which carries, tears chain information open, is passing through intermediate node along the path of reservation
When, this, which tears chain information open, will notify optical router to discharge the resource being reserved, and updates the port when unlocking output port M and averagely locks
Surely be delayed Tnew_avg；Data grouping is transmitted to destination node along the optical link made an appointment in advance；
(9) it after destination node receives data grouping, is sent in local IP kernel and handles, complete entire communication process.
2. method according to claim 1, wherein step 4a) in calculate hop count between present node and source node
Hcs, formula is as follows:
Hcs=| xc-xs|+|yc-ys|+|zc-zs|
Wherein xc、yc、zcThe X dimension coordinate, Y dimension coordinate, Z that respectively reserve resource grouping is currently located node tie up coordinate, xs、ys、
zsRespectively the X dimension coordinate of source node, Y dimension coordinate, Z tie up coordinate.
3. method according to claim 1, wherein step 4b) in calculate hop count between current node and destination node
Hcd, formula is as follows:
Hcd=| xc-xd|+|yc-yd|+|zc-zd|
Wherein xc、yc、zcThe X dimension coordinate, Y dimension coordinate, Z that respectively reserve resource grouping is currently located node tie up coordinate, xd、yd、
zdX dimension coordinate, Y dimension coordinate, the Z of respectively destination node tie up coordinate.
4. method according to claim 1, the middle average lock delay T for updating port M of step (8)new_avg, calculate public
Formula is as follows:
Tnew_avg=(1- β) (Tunlock-Tlock)+βTold_avg
Wherein, TunlockAt the time of indicating unlock output port M, TlockAt the time of locked for output port M, (Tunlock-Tlock)
Indicate that this has reserved HcdThe resource reservation of jump is grouped from the delay for requesting this port of the router to release port, Told_avg
It indicates to have been reserved H before output port McdThe average delay of the resource reservation grouping locking of jump, β indicate Told_avgWith
(Tunlock-Tlock) weight, value be 0 to 1 between.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610255814.9A CN105959239B (en) | 2016-04-22 | 2016-04-22 | The communication means of three-dimensional network on mating plate |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610255814.9A CN105959239B (en) | 2016-04-22 | 2016-04-22 | The communication means of three-dimensional network on mating plate |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105959239A CN105959239A (en) | 2016-09-21 |
| CN105959239B true CN105959239B (en) | 2018-12-21 |
Family
ID=56915231
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610255814.9A Active CN105959239B (en) | 2016-04-22 | 2016-04-22 | The communication means of three-dimensional network on mating plate |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105959239B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106936736B (en) * | 2017-04-11 | 2019-05-21 | 西安电子科技大学 | Expansible network on mating plate structure and its communication means based on bilayer layout |
| CN115276820B (en) * | 2022-07-29 | 2023-09-01 | 西安电子科技大学 | On-chip optical interconnection light source power gradient setting method using mapping assistance |
| CN115297060A (en) * | 2022-10-10 | 2022-11-04 | 荣耀终端有限公司 | Network on chip, data forwarding method and electronic equipment |
| CN117811993B (en) * | 2024-03-01 | 2024-06-07 | 山东云海国创云计算装备产业创新中心有限公司 | Three-dimensional hypercube network, routing method, device, equipment and medium thereof |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102333250A (en) * | 2011-10-26 | 2012-01-25 | 西安电子科技大学 | Low-congestion communication method and router for realizing shared path transmission of optical network on chip |
| CN102361467A (en) * | 2011-07-22 | 2012-02-22 | 西安电子科技大学 | Network structure on non-blocking optical section and communication method thereof |
| CN102413039A (en) * | 2011-10-26 | 2012-04-11 | 西安电子科技大学 | Low-blocking communication router and communication method for realizing network on optical chip |
| CN102780936A (en) * | 2012-07-17 | 2012-11-14 | 西安电子科技大学 | Optical on-chip network system of non-blocking communication and communication method thereof |
| CN105451103A (en) * | 2015-11-02 | 2016-03-30 | 西安电子科技大学 | Wavelength-allocation-based three-dimensional optical on-chip network router communication system and method |
-
2016
- 2016-04-22 CN CN201610255814.9A patent/CN105959239B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102361467A (en) * | 2011-07-22 | 2012-02-22 | 西安电子科技大学 | Network structure on non-blocking optical section and communication method thereof |
| CN102333250A (en) * | 2011-10-26 | 2012-01-25 | 西安电子科技大学 | Low-congestion communication method and router for realizing shared path transmission of optical network on chip |
| CN102413039A (en) * | 2011-10-26 | 2012-04-11 | 西安电子科技大学 | Low-blocking communication router and communication method for realizing network on optical chip |
| CN102780936A (en) * | 2012-07-17 | 2012-11-14 | 西安电子科技大学 | Optical on-chip network system of non-blocking communication and communication method thereof |
| CN105451103A (en) * | 2015-11-02 | 2016-03-30 | 西安电子科技大学 | Wavelength-allocation-based three-dimensional optical on-chip network router communication system and method |
Non-Patent Citations (2)
| Title |
|---|
| 3-D Mesh-Based Optical Network-on-Chip for Multiprocessor System-on-Chip;YE et al;《IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS》;20130430;第584-596页 * |
| AN OPTIMIZED PATH-SETUP METHOD FOR MESH BASED OPTICAL NETWORK ON CHIP;Wei Tan,Bowen Zhang,Huaxi Gu,Zheng Chen;《2015 14th International Conference on Optical Communications and Networks (ICOCN)》;20151231;第1-3页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105959239A (en) | 2016-09-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111104775B (en) | Network-on-chip topological structure and implementation method thereof | |
| CN105959239B (en) | The communication means of three-dimensional network on mating plate | |
| CN105657777B (en) | Design of Routing Protocol method based on link-quality and node transfer capability | |
| CN102238443B (en) | Method for establishing cross-domain path meeting wavelength-continuity constraints | |
| CN102377666B (en) | Flooding-based routing protocol having average-rate and burst-rate control | |
| CN103986664B (en) | A kind of mixing for network-on-chip interconnects Mesh topological structures and its routing algorithm | |
| CN103729331B (en) | Network-on-chip communication deadlock avoiding method, router and communication network | |
| CN105187313B (en) | A kind of Survey on network-on-chip topology and its adaptive routing method | |
| CN105451103B (en) | Three-dimensional light network-on-chip router communication system and method based on Wavelength Assignment | |
| Sharma et al. | A Comparative Analysis of Reliable and Congestion‐Aware Transport Layer Protocols for Wireless Sensor Networks | |
| JPH04233848A (en) | Method and apparatus for information transmission | |
| TW201145040A (en) | System and method for providing quality of service in wide area messaging fabric | |
| CN107346988A (en) | A kind of appearance based on low-track satellite network late/hold circuit network route computing method | |
| CN107396396A (en) | Support the data transfer management method of multi-source multipath | |
| CN108833293A (en) | A kind of data center's jamming control method and device based on software defined network SDN | |
| CN106254254A (en) | A kind of network-on-chip communication means based on Mesh topological structure | |
| Li et al. | Connection-oriented and connectionless remote entanglement distribution strategies in quantum networks | |
| CN106209627A (en) | Vehicle-mounted distributing based on node-node transmission ability focuses on Routing Protocol method | |
| CN103888976A (en) | Link selection method combined with network scheduling and routing | |
| CN109525304A (en) | Perceptual computing stores the integrated space intelligent network architecture | |
| CN103916253A (en) | Information transmission method and system on basis of information center network | |
| CN104994508A (en) | Cognitive radio mesh network resource distribution and routing method | |
| KR101465498B1 (en) | Apparatus and method for adaptive routing of hybrid optical networks on chip and hybrid optical networks on chip system using the same | |
| CN101436980A (en) | Generation of data packet forwarding table by flooding method | |
| US20120230330A1 (en) | Method for controlling area boundary, method and system for establishing connection in multilayer network |
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 |