CN101155121A - Switching method and node structure supporting burst data packets and IP packets - Google Patents

Abstract

The invention provides an all-optical switching method and a node structure supporting two granularities of burst data packet/IP grouping. It uses electrical cache instead of optical cache, which is currently difficult to implement, and utilizes the "best effort" transmission feature of IP to store IP packets transmitted between adjacent nodes and/or some IP packets with low QoS requirements in electrical caches. According to the information of network resources reserved by the burst control packet, these IP packets are inserted into the gaps between the burst data packets as much as possible for transmission. That is, IP packets pass through OBS nodes like an hourglass, while burst data packets are transparently transmitted in the optical domain. The method can effectively reduce the number of burst data packets in the network, and the data can be evenly distributed on the wavelength channel by using the electric random access memory. It can not only provide efficient statistical multiplexing capability, but also greatly reduce the blocking rate and data loss rate of the network, and it is easy to implement.

Description

Switching method and node structure supporting burst data packets and IP packets

technical field

The invention relates to the technical field of all-optical communication, in particular to an all-optical switching method and node structure supporting multiple granularities.

Background technique

In recent years, with the rapid development of the Internet and people's demand for multimedia information, technologies such as network transmission and exchange are facing severe challenges. How to effectively meet this growing bandwidth demand has become one of the main topics in network research. Wavelength division multiplexing (WDM) technology provides massive transmission bandwidth for optical fiber communication. One optical fiber carries hundreds of wavelength channels, and the transmission bandwidth reaches tens to hundreds of terabits per second. The currently used core routers receive high-speed optical signals, process them in parallel after optical/electrical conversion, and then multiplex them into high-speed optical signals for forward transmission through electrical/optical conversion. In order to maintain wire-speed packet forwarding, the core router not only needs a large number of optical/electrical/optical conversion devices and a large number of parallel processing, but also requires a large number of multiplexing and demultiplexing. Due to the intrinsic characteristics of electrons, the processing capacity and switching speed of the switching module are restricted. The development of the current electronic exchange and information processing network is close to the limit of the electronic rate, and its inherent shortcomings such as RC parameters, time jitter, drift, crosstalk, and response speed limit the improvement of the exchange rate, which is the "electronic bottleneck". Due to the existence of "electronic bottleneck", the switching rate is restricted by the rate of the core router's backplane, and the structure is complex and expensive. The way to overcome the "electronic bottleneck" is to perform all-optical switching directly.

All-optical switching technology refers to directly switching the input optical signal data to different output ports of the router without any optical/electrical and electrical/optical conversion. The all-optical switching technologies under research mainly include: Optical Circuit Switching (OCS), Optical Burst Switching (OBS) and Optical Packet Switching (OPS). Researchers have compared three optical switching technologies, among which optical burst switching has achieved a good balance between optical circuit switching and optical packet switching. It combines the advantages of light and electricity. Its switching granularity is moderate, it can not rely on optical cache, it is easy to implement, and its bandwidth utilization rate is high. It can meet the requirements of network business growth and business diversity. The switching time adopts nanosecond level according to the granularity. Or implemented by microsecond-level optical devices, it can statistically multiplex optical network bandwidth resources.

Optical burst switching is a kind of all-optical switching with sub-wavelength granularity. By encapsulating user data into burst data packets of a certain length according to a certain aggregation method (the duration is from hundreds of microseconds to tens of milliseconds at a rate of 10Gbps) transmission. This greatly reduces the requirements on the processing speed of electronic devices and the speed of optical switches. At the same time, because the burst data packet of optical burst switching is a super long data packet, the number of burst control packets is much smaller than that of optical packet switching, and the control information needs much less optical/electrical and electrical/optical conversion at the OBS node. , data read and write operations are less, so it is easier to implement than optical packet switching. Compared with OPS, OBS has lower requirements on optical devices and is subject to fewer restrictions; compared with OCS, OBS has a higher degree of statistical multiplexing capability, and is more suitable for bursty Internet network services.

The core idea of OBS technology is to try to greatly reduce the switching granularity of the optical network, provide efficient statistical multiplexing capabilities, and fully consider the limitations of optical devices to reduce the requirements for optical devices. In the OBS network, the allocation of resources takes the duration of the burst data packet as the basic granularity, which makes the optical switching rate of the OBS node can be greatly reduced. But contention occurs when two or more burst data packets are required to be sent simultaneously from the same port, on the same wavelength, of an OBS node. Since there is no optical random access memory (ORAM) similar to electrical random access memory (RAM), the ability of OBS nodes to deal with burst data packet collisions is greatly restricted. The currently researched technologies to solve the contention include: using fiber delay lines (FDLs) at the input burst data packet to realize a small amount of optical buffer, deflection routing for burst data packet, wavelength conversion and burst data segmentation, etc. These technologies are to divert one or some of the burst data packets to another wavelength, port or optical fiber in the OBS node when the burst data packets compete. They can avoid the conflict at the OBS node at the current moment, but will cause new competition with a high probability. At the same time, implementing a small number of optical buffers using fiber delay lines cannot effectively solve the contention problem under high load conditions. Offset routing is easy to implement, but limited by the network scale and its connectivity, it is unfavorable to provide ideal network performance. Wavelength conversion can reduce the average delay and data loss rate, but the full-wavelength conversion technology is still immature, and the price of partial wavelength conversion is relatively expensive. When the network load is high, the use of partial wavelength conversion has limited improvement in network performance. Burst data segmentation can reduce the data loss rate, but it increases the difficulty of realizing the optical burst switching network. In order to resolve conflicts, it is also necessary to use complex switching scheduling algorithms and conflict resolution strategies in the control module of the OBS node. At the same time, the cache capacity and aggregation algorithm of the aggregation module must be carefully designed. This will increase the implementation complexity of the node.

How to optimize the switching mode and resource allocation mode inside the nodes of the all-optical network according to the needs of network services is an important research topic. That is, to study an all-optical switching method and network node structure that supports multiple granularities, and has a simple and economical node structure, is still a hot issue worthy of research.

Contents of the invention

For above-mentioned analysis to the method for current all-optical switching and solution data competition, the purpose of the present invention is to provide a kind of new switching method for all-optical communication network, try one's best to provide efficient statistical multiplexing ability, fully consider simultaneously The limitation of optical devices reduces the requirements for optical devices.

Technical scheme of the present invention is as follows:

An all-optical communication switching method supporting burst data packets and IP packets, comprising:

After the data from the traditional network interface arrives at the OBS node, the data transmitted between the adjacent nodes (if the source node and the sink node pair are nodes connected through the OBS link, such nodes are called adjacent nodes) use IP packets to transmit data on the OBS link. On the road transmission, the data transmitted between remote nodes (if the source and sink nodes are not directly connected through the OBS link are called remote nodes), the data transmitted between them is aggregated into a burst data packet and transmitted on the OBS link.

After the above-mentioned IP packet data and burst data packets enter the sending cache of the OBS node, the OBS node first reserves resources for the passing burst data packets, and then arranges the sending node to go on the road to the remote node according to the reservation status of the local resources. At the same time, use the idle time gap of the uplink wavelength channel of this node to send IP packet data.

For data with low QoS requirements, although these data are transmitted between remote nodes, they may not be aggregated, and IP packets are used to transmit them on the OBS link.

A node structure supporting burst data packets and IP packet switching, mainly composed of a local add/drop module and an all-optical switch module, the local add/drop module includes a convergence/de-convergence module and an IP packet sending/receiving module. Among them, the aggregation/de-aggregation module is used to aggregate data into burst data packets when sending data, and perform de-aggregation function when receiving data; the IP packet sending/receiving module is used for transmission and reception of IP packet data. The all-optical switching module based on electrical control is used to switch the local on-road burst data packets, IP packets and passing burst data packets to corresponding output wavelengths according to wavelength resource reservation.

The OBS node stores the on-road IP packets and burst data packets in the local ERAM, and schedules these on-road data through a TS (Traffic Spacing) scheduler.

It further includes a service classifier, which is used to classify the service data arriving at the node according to the destination address or QoS requirements, that is, the data with high QoS requirements whose destination address is the remote node enters the aggregation/disassembly module; the destination address is the adjacent node Or data with low QoS requirements enters the IP packet sending/receiving module.

The exchange method for supporting burst data packets and IP packets includes two layers of content: first, in terms of data packets, the data transmitted between adjacent nodes is not aggregated to form burst packets, and IP packets are directly used on the OBS link Transmission: Data transmission between remote nodes can be treated separately according to QoS requirements. Data with high QoS requirements needs to be aggregated into burst data packets for transmission on the OBS link, while data with low QoS requirements can be composed without aggregation Burst data packets are directly transmitted on the OBS link with IP packets. Secondly, in terms of sending data, the OBS node stores the passing IP packets and local on-road data (including burst data packets and IP packets) in the ERAM of the node, and uses the TS scheduler to store the data stored in the local ERAM. The data is scheduled; the OBS node processes the control information packet and reserves the wavelength resource for the passing burst data packet. According to the reservation situation of the local resource, it arranges the sending node to add the burst data packet to the remote node, and at the same time uses the wavelength of the local node. The channel's idle time slots send IP packet data.

The exchange method supporting burst data packets/IP packets proposed by the present invention utilizes IP packets to fill the gaps between burst data packets, which not only reduces the number of burst data packets in the network, but also makes use of ERAM to make data in the wavelength channel Evenly distributed on the network, not only can provide efficient statistical multiplexing capability, but also can greatly reduce the blocking rate and data loss rate of the network. At this time, the OBS node needs to be equipped with more ERAMs, no additional devices are needed, and the implementation is simple.

Description of drawings

Fig. 1 is a schematic diagram of a switching node structure supporting burst data packets and IP packets of the present invention.

Fig. 2 is a schematic diagram of the all-optical switching method supporting burst packets/IP packets of the present invention.

Fig. 3 is a schematic diagram of sending data by a switching node in the present invention.

Fig. 4 is a schematic diagram of sending IP packets using the gap between burst data packets.

Figure 5 is a linear network topology diagram composed of four nodes.

FIG. 6 shows the packet loss rate performance of the network topology shown in FIG. 5 obtained by the switching method supporting burst data packets and IP packets at two granularities and switching node structure simulation according to the present invention.

in:

1——Wavelength Division Multiplexer 2——Wavelength Division Multiplexer

Detailed ways

The present invention will be described in detail below in conjunction with the accompanying drawings, but the scope of the present invention is not limited in any way.

A schematic diagram of the node structure of the switching method supporting burst data packets/IP packets proposed by the present invention is shown in FIG. 1 . The OBS node shown in the figure does not divide into two parts, the edge and the core. It consists of a service classifier, a control module, an aggregation/de-aggregation module, an IP packet sending/receiving module, a data storage, a TS scheduler, an optical sending/receiving module, and an all-optical switching module. module, optical wavelength division multiplexer/demultiplexer, etc.

When the data with low QoS requirements is not considered to be transmitted in IP packets, after the data from the traditional network interface arrives at the OBS node, it first passes through the service classifier, and the service classifier classifies the data with the destination address as the adjacent node and the data with the destination address as the remote node. The data is classified, and the data whose destination address is the remote node enters the aggregation/de-aggregation module, and then encapsulates the burst data packet according to a certain encapsulation algorithm; the data whose destination address is the adjacent node enters the IP packet sending/receiving module. The local on-road burst data packets and IP packets are stored in the ERAM of the node, and the on-road data is scheduled through the TS scheduler, and the nodes reserve resources for the passing burst data packets as shown in Figure 4. After the resource request is reserved by the burst data packet, according to the local resource situation, the sending node is arranged to send the burst data packet to the remote node, and at the same time, the IP packet data is sent by using the idle time gap of the wavelength channel of the local node. After the data is sent by the optical sending module, it enters the all-optical switching module and enters the link through the wavelength division multiplexer. When receiving data, the data from the link first passes through the optical wave decomposition multiplexer, and then the all-optical switching module switches the data to the corresponding output port according to the destination address. The data whose destination address is the node is received by the optical receiving module and cached in the electrical memory, and then sent to the aggregation/de-aggregation module and the IP packet sending/receiving module respectively through the scheduling of the control module, and finally classified by the service classifier through the traditional The network interface reaches the destination; the data whose destination address is other nodes is switched to the output port by the all-optical switch module, and then enters the next-hop link through the optical wavelength division multiplexer for transparent transmission to the destination.

A schematic diagram of sending data by a switching node supporting two granularities of burst data packets/IP packets proposed by the present invention is shown in FIG. 3 . After the data from the traditional network (such as Ethernet, IP, ATM, SDH network) arrives at the node, it needs to be processed separately by the service classifier according to the destination address, where the data whose destination address is the remote node is encapsulated by the convergence adaptation module Enter the sending buffer, and the data whose destination address is the adjacent node directly enters the sending buffer. The TS scheduler schedules the transmission of the data in the transmission buffer according to the local resource reservation situation. The scheduled burst data packets or IP packets enter the all-optical switching module after electro-optical conversion, and are switched to the output port according to the routing information. The WDM enters the link for transmission to the next hop. The passing burst data packets are transparently transmitted in the optical domain.

In the design of switching node structures with two granularities supporting burst data packets/IP packets, service classifier, aggregation/de-aggregation module, control module, memory for adding and dropping data, TS scheduler, optical sending/receiving module , an all-optical switching module composed of high-speed optical switches, optical wavelength division multiplexers/light wavelength division multiplexers, etc. are the basic devices and modules for building nodes. As an example, by selecting these basic devices and modules listed in the present invention, according to the switching method supporting burst packets/IP packets shown in FIG. 2, the switching node shown in FIG. 1 can be formed. By means of OPNET ^TM simulation software, according to the node structure and switching method proposed by the present invention, a 4-node optical burst switching network as shown in Fig. 5 is established. Fig. 6 is the packet loss rate performance of the network obtained by simulation. Each node in the network has four traditional network interfaces to inject data into the network, and it is assumed that these interfaces are connected with a large number of independent Poisson traffic sources. The grouping algorithm at the node is a time-length product threshold algorithm, and the threshold is 9 megabit·millisecond (Mbms). The routing protocol adopts the static path selection method shown in Figure 5, and the wavelength allocation algorithm adopts the first fit algorithm.

The curves in Figure 6 are all using the JET (Just Enough Time) resource reservation protocol. The curve marked "No IP packet insertion" in the figure indicates that all data is aggregated and assembled into burst packets at the node, and IP is not used. Packets fill the gaps between burst data packets; the curve marked "with IP packet insertion" in the figure indicates that the node sends IP packets to fill the gaps between burst data packets for data between adjacent nodes. It can be seen from Fig. 6 that the average data loss rate represented by the curve "with IP packet interleaving" is about 50% lower than that indicated by the curve "no IP packet interleave".

The above are the embodiments of the present invention. According to the disclosed content of the present invention, some similarities and alternatives that those skilled in the art can obviously think of should fall into the protection scope of the present invention.

Claims (6)

1. An all-optical communication switching method that supports burst packets and IP packets, comprising: after the data from the traditional network interface arrives at the OBS node, the data transmitted between adjacent nodes is transmitted on the OBS link using IP packets, Other data are aggregated into burst data packets and transmitted on the OBS link. 2. the all-optical communication exchange method that supports burst data packet and IP grouping as claimed in claim 1, is characterized in that: OBS node at first reserves resource to give the burst data packet of crossing, according to the reservation situation of local resource, then Arranging the sending node to add the burst data packet to the remote node, and at the same time send the above IP packet data by utilizing the idle time gap of the on-channel wavelength channel of the node. 3. The all-optical communication switching method supporting burst data packet and IP grouping as claimed in claim 1 or 2, characterized in that: the data with low QoS requirements is not aggregated, and IP grouping is used to transmit on the OBS link . 4. A switching node structure supporting two granularities of burst data packet/IP grouping, consisting of a local add/drop module and an all-optical switch module, characterized in that: the local add/drop module includes a convergence/disassembly module and IP packet sending/receiving module, wherein the aggregation/de-aggregation module is used to aggregate data into burst data packets when sending data, and perform de-aggregation function when receiving data; IP packet sending/receiving module is used for IP packet data Transmission and reception: an all-optical switching module, used to switch local on-road burst data packets, IP packets, and passing burst data packets to corresponding output wavelengths according to resource reservations. 5. the switch node structure of two kinds of granularity that supports burst data packet/IP grouping as claimed in claim 4, it is characterized in that: node stores IP grouping and data on the local road in the ERAM of node, through a The TS scheduler schedules the data stored in the ERAM. 6. The switch node structure supporting two kinds of granularity of burst data packet/IP grouping as claimed in claim 4 or 5, it is characterized in that: it further comprises a service classifier, which is used to sort the service data of the arriving node according to the destination address Or QoS requirements are classified, that is, the data whose destination address is a remote node with high QoS requirements enters the aggregation/disassembly module; the destination address is an adjacent node or data with low QoS requirements enters the IP packet sending/receiving module.

Images