CN103532653A - Segmented recombination method used for Gigabit passive optical network - Google Patents

The invention discloses a segmented recombination method used for a Gigabit passive optical network. A plurality of parallel flow lines are reduced into one. The method comprises the following steps that a, the user data length is monitored in real time; b, the user data is padded according to the system data bus width, so the user data realizes the width alignment with a system data bus; c, finally, the client data is separated into a GEM (Gigabit passive optical network encapsulation method) effective load, an optical network unit uses an instruction level parallelization and task level parallelization processing method, the service data realizes the mutual communication through a passage or a FIFO (first in first out) queue, and in addition, the loop level flow line and the iterative module dispatching are deployed. The segmented recombination method used for the Gigabit passive optical network provided by the invention has the advantages that the higher transmission speed can be realized, the hardware resource use and development period can be simplified, and in addition, the system performance cannot be influenced.

Segmentation recombination method for gigabit passive optical network

Technical field

The present invention relates to a kind of packet segmentation recombination method, relate in particular to a kind of segmentation recombination method for gigabit passive optical network.

Background technology

One of Optical Access Network technology that gigabit passive optical network (GPON) is current popular.GPON is used is single point-to-point configuration framework and single network topology is provided, and by means of transmission container (TCONT) and Dynamic Bandwidth Allocation (DBA), provides data, voice and the Video service with QoS.

GPON can support TDM service and the ever-increasing application of needing bandwidth badly of old edition, such as the high definition TV via IP.It can also be seen as obtain extensively approval for implementing the candidate scheme of access (NGA) network of future generation, the upstream that can supply the downstream of 10 gigabit/sec and be less than or equal to this speed.Implementing the technological challenge that 10 gigabit/sec must overcome is resolved.Except supplying reliable and high-quality access network service, realize cost embodiment to one's profit and remain a major challenge that industry faces.

Summary of the invention

Technical problem to be solved by this invention is to provide a kind of segmentation recombination method for gigabit passive optical network, can realize more high transfer rate, and simplifies hardware resource and use and the R&D cycle, and can not affect systematic function.

The present invention solves the problems of the technologies described above the technical scheme adopting to be to provide a kind of segmentation recombination method for gigabit passive optical network, and many parallel pipelines are reduced to one, comprises the steps: a) Real-Time Monitoring User Data Length; B) according to system data bus width, fill user data, user data is alignd with system data bus width; C) finally client data is separated into GEM pay(useful) load.

The above-mentioned segmentation recombination method for gigabit passive optical network, wherein, described optical network unit is used instruction level parallelism and task level parallelization processing method, described business datum communicates with one another by passage or First Input First Output, and disposes the module schedules of loop level production line and iteration.

The present invention contrasts prior art following beneficial effect: the segmentation recombination method for gigabit passive optical network provided by the invention, by two FPGA plates structure optical network units and optical line terminal, carry out complete or collected works' simulation, the segmentation of filling while avoiding parallelization pipeline processes by packet is recombinated, many parallel pipelines are reduced to one, thereby realize more high transfer rate, and simplify hardware resource and use and the R&D cycle, and can not affect systematic function.

Accompanying drawing explanation

Fig. 1 is that the present invention is for the segmentation recombination method structural representation of gigabit passive optical network;

Fig. 2 is streamline in gigabit passive optical network and the use rough schematic view of concurrency;

Fig. 3 is the slot offset FSM schematic diagram of the segmentation restructuring of gigabit passive optical network.

Embodiment

Below in conjunction with drawings and Examples, the invention will be further described.

Fig. 1 is that the present invention is for the segmentation recombination method structural representation of gigabit passive optical network.

Refer to Fig. 1, the segmentation recombination method for gigabit passive optical network provided by the invention, adopts two FPGA plates to be connected and forms an optical network unit; Adopt two FPGA plates to be connected and form an optical line terminal; Message transmission rate between described two FPGA plates is 1.24 gigabit/sec; A plurality of optical network units and optical line terminal are connected to form to GPON system by monomode fiber and passive optical splitters; Described FPGA plate is all connected to ethernet client, for receiving data, voice and video source, and produces the upstream business with different staggered and different frame signs, with simulated light network element; Described optical network unit is used instruction level parallelism and task level parallelization processing method, and described business datum communicates with one another by passage or First Input First Output, and disposes the module schedules of loop level production line and iteration; The present invention adopts kilomega passive optical network packaging method (GEM) frame structure, comprising: GEM frame head part and GEM frame pay(useful) load part, many parallel pipelines are reduced to one, and specifically comprise the steps:

Step S1: Real-Time Monitoring User Data Length;

Step S2: according to system data bus width, fill (padding) user data, user data is alignd with system data bus width;

Step S3: finally client data is separated into GEM pay(useful) load.

The concurrency of using in GPON simulator and the common technique of streamline are as follows: in this GPON simulator, used instruction level parallelism (instruction level parallelization) and task level parallelization (task level parallelization).The former belong to other parallelization of lowermost level, and the latter is logic-based based on space, and wherein task communicates with one another by passage or first-in first-out (FIFO).Disposed thick and fast the module schedules of loop level production line and iteration, these loop level production lines started a new loop iteration before stopping at current loop.As shown in Figure 2, this streamline has the delay of four clock circulations, make when separator is when preparing to need " current " data of transmission, RocketIO in fact sends the data in " past ", and the data of being somebody's turn to do " past " are delayed in this streamline.From another side, when RocketIO sends " current " data, this separator must calculate and prepare " future " data that will send after four clock circulations.

The present invention is by the explanation of global function GPON simulator, and it can tackle different bag skews in the timesharing of splitting of implementing by GPON standard (G.984.3) defined.This GPON simulator need to compare with proposed segmentation recombination method.The embodiment of standard compliant design has shown needed substantive work, this may form most of cost of business development, in addition, intensive streamline and concurrency will make area and power affect more the choice between area, power and clock rate.Finally, this complexity has produced difficulty, and difficulty is that in the real-time constraint condition meeting in design, implementing High Speed I/O system supports 10 gigabit/sec data flow.

Depend on actual needs, have two subject matters in segmentation recombination method flow process, these two subject matters have affected the enforcement of concurrency and streamline, are offset (offset) and split minute (fragmentation).First problem (skew) be by with as when the hardware data width being used from serial-to-parallel conversion and streamline is subsequently compared, the different byte offset of the data that receive continuously causes.When ethernet frame is considered to represent the client data of arrival, separator will distribute time slot according to the length of each data field.If the time slot distributing is aimed at data-bus width, slot offset will be always zero so, otherwise it will be any one in the x shown in equation 1.W _dit is the data width in bit.

Second Problem (split minute) is that the demand by network efficiency causes, that is people can not only be wasted in remaining space in frame and wait for next frame rather than split minute.Therefore, split divided by for express network being a feature of wanting.In order to hold the frame arriving with the different different time-gap of length, GPON embodiment need to have x the streamline copying.In the situation of this simulator, x=5, because the data-bus width in RocketIO is five bytes.In addition, this embodiment must be included the appearance of the data of five streamlines between any two segmentations to consider in.Fig. 3 a represents the variation of time slot skew (offset) state machine in existing manner of execution, and after Fig. 3 b represents to use the present invention, time slot is offset (offset) state machine.The complete logic embodiment of each streamline for can infinitely moving, the transition (comprising self, complete mesh network) of each from each state to five state of existence.Each state is corresponding to a streamline; No matter when new data arrive, new skew need to be calculated based on current skew.Therefore the FPGA resource of, using sharply expands; Sequential between gate also becomes and is difficult to meet, and has formed the obstacle of 10 gigabit/sec NGA.

The GPON segmentation recombination method of novelty provided by the invention is reduced to only one by five streamlines that copy, and this just means a unified slot offset.First to client data, (padding) filled in application, then client data is separated into GEM pay(useful) load.Under new segmentation recombination method, the length of client data is monitored, and this separator will be always align ethernet frame (via filling) with data-bus width, therefore only need a streamline to hold the ethernet frame of any length and can between different slot offset, not slide (slippage).This will reduce large width-radius ratio compared with the number of device, and the latter can cause considerable delay in FPGA network.And it also implements to split minute, makes proposed segmentation recombination method equally efficient with standard GPON.But can greatly reduce Time To Market and R&D costs.In addition, it can also reduce the required FPGA resource of identical function.

The trigger that the novelty of the key performance attribute that uses Xilinx ISE and this design when location and route (place-and-routed) and standard segmentation recombination method embodiment are used and look-up table (LUT) have nearly reduced 80%.The time sequence improving of reaching four times; As everyone knows, in practice, when Timing Constraints reaches capacity, even be all very difficult by 1 nanosecond of time sequence improving.But new segmentation recombination method was but improved for 24 nanoseconds.This method is different from many solutions by sacrificing the method for amassing to obtain higher clock speed.(for example, can insert pipeline stages increases clock rate, because reduce the number of the operation independent in a clock circulation, can increase attainable speed).Thereby this is a good solution, it has used less area, has clock speed faster, still keeps identical functional simultaneously.Therefore the power, consuming also can reduce.Most important benefit is to dispose in practice the crucial motive force of this technology in the temporal remarkable minimizing of research and development (using the segmentation recombination method proposing to estimate approximately can reduce 70%) ，Zhe Shi manufacturers.The GPON that all these benefits make to be enhanced becomes the powerful candidate scheme of 10 gigabit/sec NGA.

Same benefits is also applicable to receiver part, because restructuring will not maintain the different time-gap skew of five streamlines that copy, does not include in and considers; It just treats the ethernet frame that any length is not identical coequally, and will only need a unique streamline to carry out recombination data.User Network Interface (UNI) can recover received data by gap information filtering and according to this gap information, and is transferred to client.

If split the uncared-for words of minute function, the needs of calculated data skew will do not there are not so; Skew will be always 0, and segmentation recombination method is simplified.Also may query offset alignment (filling) and will cause not good efficiency.Below displaying being split minute is an important feature, can improve efficiency considerablely, and fill, only can in very little, acceptable degree, affect efficiency.

The odds for effectiveness of verifying below by OPNET analog result:

About which kind of Distribution Statistics when Network is carried out to modelling, be optimum, had a large amount of arguements.When to traffic performance modelling, be once widely used Poisson process (Poission process), afterwards researchers whether argue Network be self similarity.Currently also exist a kind of viewpoint to think, when many self-similar traffics flock together, exponential distribution is followed in Network.But arguement Internet service model is not object of the present invention.No matter this problem how, the GPON segmentation recombination method proposing shows the excellent performance that is applicable to various types of traffic.

Suppose that network during normal use has 0.7 load.The expense factor 0.06 that GPON is general is not considered, and has only considered that the segmentation recombination method efficiency proposing makes its impact more clear.Bag size is followed exponential distribution.Average packet size comprises from 100 bytes to 1500 bytes, and step size is 200 bytes.Wrapping big or small Hurst parameter is 0.7, and the fractal time started, (Fractal Onset Time Scale) was 1.0, is suitable for self similarity and distributes.In order to verify the extensibility for the novel segmentation recombination method of NGA, GPON capacity is modeled as from 1.25 gigabit/sec to up to 10 gigabit/sec.Suppose that the GPON (but have to split minute not have and fill) of use standard segmentation recombination method has the efficiency (efficiency of unity) of homogeneous, can be via the efficiency of the GPON of simulation new segmentation recombination method of research use in bag size and load (have split minute also have filling).Also there are other two kinds of possible options, i.e. " without splitting minute also without filling, " and " nothing split minute but have filling ", these two kinds of options also simulated to compare.Finally, real the Internet trace files (internet trace file) can be verified as source data the correctness of above-mentioned simulation.In the present invention, bandwidth efficiency E is defined as follows:

Lte is the total length of ethernet frame, the total length of pt for filling, nf _wfor owing to not splitting the time slot altogether wasted in minute in place.GPON segmentation recombination method proposed by the invention always wraps size simultaneously and is also increasing increasing, no matter the most why the type of service of using and GPON hold.The minimum efficiency of the solution proposing is 0.97 and tends to homogeneous along with bag size increases.If omit and to split minute, without filling general trend, be decrease in efficiency.Obviously, without splitting minute but there is the efficiency of filling minimum.When bag size is greater than 200 byte, the segmentation recombination method solution proposing can also other two kinds of options of extraordinary execution.When bag size is very little, the number comparability of the byte being filled, in original length, is therefore filled and has been dominated the impact on efficiency.When bag size increases, in being greater than the situation of 200 bytes, splitting and minute start dominate and affect efficiency in a big way.When each ethernet frame size is very little, a GPON frame contains more ethernet frame and is filled more continually so.When ethernet frame size is enough greatly time, the number of the ethernet frame in a GPON frame becomes less, and if the probability that they are filled so can lowlyer not exist any split minute, last part will slattern time slot.And can find with the curve of circle with foursquare curve intersection point---when GPON capacity increases, this intersection point will move to the right side of x axle.This is that GPON frame sign can increase (1.25 gigabit/sec are that 19440 bytes, 2.5 gigabit/sec are 38880 bytes, and 10 gigabit/sec are 155520 bytes) linearly because when GPON capacity increases, and then filling will be leading increasedly gradually.

In sum, from the common checking of hardware GPON simulator and software OPNET simulation, can find out, compare to the solution proposing in GPON standard, the GPON segmentation recombination method proposing has greatly been simplified enforcement.Specifically, it should be noted that, it has taken nearly 80% FPGA resource, has reached the hardware clock speed that speeds 4 times, and according to author's experience herein, by research and development time decreased reach 70%.Therefore, this scheme can be used as the candidate scheme of putting into practice of 10 gigabit/sec NGA.Use simulation to show, the segmentation recombination method proposing has retained the important minute feature of splitting, and efficiency is almost identical with standard GPON.For different service distribution, packet size and business load, there is robustness, be suitable for the GPON of the friction speed between 1.25 gigabit/sec and 10 gigabit/sec.

Although the present invention discloses as above with preferred embodiment; so it is not in order to limit the present invention, any those skilled in the art, without departing from the spirit and scope of the present invention; when doing a little modification and perfect, so protection scope of the present invention is worked as with being as the criterion that claims were defined.