KR20090022473A - Forwarding table for layer 3 switching - Google Patents

Abstract

A forwarding table for 3-layer switching is provided to overcome the restriction that the big size of hardware LPM table must be equipped, thereby realizing a cheap switching device having excellent performance. A forwarding table for 3-layer switching comprises an FIB(400), an LPM table(410), a management unit(420) and a forwarding unit(430). In the LPM table, entries of the FIB in which the possibility of being used is high are stored through the control of the management unit. The entries having the high possibility are determined by monitoring the number of times entries are used for forwarding. The management unit selects the entries which have the high possibility of being used and does not exist in the LPM table, and then inserts the selected entries into the LPM table.

Description

Forwarding table for layer 3 switching}

1 is a diagram conceptually illustrating a relationship between a routing protocol, a RIB, an FIB, and an LPM table.

2A and 2B are block diagrams illustrating the configuration of a device having a three-layer switching function.

3 is a view for explaining the concept of using the LPM table as a cache (cache) according to an embodiment of the present invention.

4 is a block diagram illustrating an apparatus having a three-layer switching function according to an embodiment of the present invention.

5A and 5B are block diagrams illustrating an LPM table used according to an embodiment of the present invention.

6 is a diagram illustrating a process of periodically monitoring a usage field according to an embodiment of the present invention.

FIG. 7 is a diagram illustrating an LPM table in which a hit bit is recorded and used as a cache according to an embodiment of the present invention.

8 illustrates an LPM tree for explaining a bundle processing process according to an embodiment of the present invention.

9A and 9B are diagrams for describing a concept of performing packet forwarding using an LPM table to which a representative entry is applied according to the representative entry using method according to the first embodiment of the present invention.

The present invention relates to layer 3 switching technology, and more particularly, to a forwarding table for efficiently performing layer 3 switching.

Layer 3 switching means forwarding (delivering) a packet from one network to the next network (i.e. next hop) on the third layer of Open Systems Interconnection (OSI) to reach its final destination. Examples of devices having a three-layer switching function include a three-layer switch, a router, and an optical network unit (ONU) of an Ethernet-Passive Optical Network (E-PON). . In this section, a device having a three-layer switching function will be described as a router for convenience.

In order for a router to perform three-layer switching, routing information such as information about a route reaching each network, a metric (eg, hop count, cost, etc.) of the route must be obtained in advance. As a method of obtaining such routing information, there are a static routing method directly input by an administrator and a dynamic routing method obtained through a routing protocol for exchanging routing information between routers. Examples of routing protocols include Routing Information Protocol (RIP), Open Shortest Path First (OSPF) protocol, and Border Gateway Protocol (BGP).

Under the dynamic routing environment, the router's dynamic routing daemon establishes a RIB that contains routing information obtained through the above-described exchange as an entry. The router extracts the best entry in terms of metrics from among the entries of the RIB to construct the FIB. That is, only routes selected as valid in the RIB are included in the FIB as entries. Specifically, the FIB includes entries that bind each network address, a corresponding output port of the router, router information of the next hop, and the like.

On the other hand, a switching scheme using FIB (i.e., a software-based switching scheme) is possible, but a hardware-based switching scheme using a longest prefix matching (LPM) table has been introduced for faster searching. .

According to the hardware-based switching scheme, some of the entries of the FIB look up the stored LPM table to perform switching. This search uses the LPM algorithm, which selects among the entries in the FIB that have the most matching network address in the order of the higher bits of the destination address of the incoming packet to determine the next hop. to be. On the other hand, a default entry such as 0.0.0.0/0 is provided in case there is no entry in the LPM table that matches the destination address of the received packet. In this case, packet forwarding takes place to the next hop recorded in the default entry.

The conceptual relationship between the routing protocol, the RIB, the FIB, and the LPM table as described above is illustrated in FIG. 1.

2A and 2B are block diagrams illustrating the configuration of a device having a three-layer switching function.

The apparatus 200 of FIG. 2A is described below. The CPU 205 performs overall control of the apparatus 200, performs RIB establishment by the dynamic routing daemon, and FIB establishment, and stores the same in the first repository 210. An example of the first reservoir is an inexpensive and high capacity SDRAM.

The switching chip 215 receives the entries of the FIB from the CPU 205, constructs an LPM table in the second repository 220, searches the constructed LPM table, and forwards the received packet to the next hop. That is, the received packet is input to the switching chip 215 through any one of the four ports (225, 230, 235, 240), and is output through any one.

Functions of the CPU 255, the first reservoir 260, the switching chip 265, the second reservoir 270, and the ports 275, 280, 285 and 290 included in the device 250 of FIG. It is the same as the corresponding structure of FIG. 2A, except that the first reservoir 270 is included in the switching chip 265 and constitutes one chip.

1, 2A and 2B, it can be seen that the RIB and the FIB are built in software through the CPUs 205 and 255, and the LPM table is built in hardware through the switching chips 215 and 265. . It can also be seen that the switching chips 215 and 265 need to be assisted by the CPUs 205 and 255 to retrieve the FIB while the LPM table is directly accessible.

On the other hand, since the RIB includes tens of thousands to hundreds of thousands of entries according to a routing protocol, the RIB includes at least tens of thousands of entries even if the FIB is constructed by collecting / selecting them. However, a problem that arises in reality occurs in the process of including tens of thousands of entries of the FIB in the LPM table, that is, the second repository (220, 270).

For high-speed switching, the second reservoirs 220 and 270 must support much faster access speeds than the first storage elements 210 and 260, which are generally in the form of SDRAM, and in some cases, content addressable memory (CAM) and the like. Since they must have the same unique structure, there are limitations in terms of price and storage capacity. Therefore, in the commercial apparatus, the number of entries included in the LPM table is limited to tens of thousands or less, and as a result, a problem arises that it is impossible to include all the entries of the FIB in the LPM table.

Conventional methods for solving this problem include a method of storing entries of the FIB only as long as the LPM table can be supported, a method having second storage elements 220 and 270 capable of supporting a sufficient LPM table, and a capacity of the LPM table. There was a way to limit the network configuration. However, in the first method, some entries are not reflected in the LPM table, causing false packet forwarding, the second method is disadvantageous in terms of device cost, and the third method is a limitation in configuring a desired network.

Therefore, in order to solve this problem, a method for efficiently performing three-layer switching under the hardware constraints described above is required.

An object of the present invention is to provide a forwarding table for efficiently performing three-layer switching.

In order to achieve the above technical problem, a first aspect of the present invention provides a forwarding table for three-layer switching including at least one entry in which a field value is recorded to refer to a separate forwarding table.

In order to achieve the above technical problem, a second aspect of the present invention provides a network address field; Next hop field; And a field indicating whether the network address field is used, and at least one of the next hop fields provides a forwarding table having a value for referring to external forward information.

Hereinafter, embodiments of the present invention for solving the above technical problem are described with reference to the accompanying drawings.

3 is a view for explaining the concept of using the LPM table as a cache (cache) according to an embodiment of the present invention, in more detail the concept of managing the LPM table to include the maximum number of active entries in the LPM table; It is a figure for demonstrating.

Referring to FIG. 3, it can be seen that an active entry is additionally stored in the LPM table from the FIB, and an inactive entry is deleted from the LPM table.

In this case, the active entry means an entry having high availability, for example, a hit probability. A highly available entry means an entry that is in use or expected to be used for packet forwarding. An inactive entry is an entry whose attributes are opposite to those of the above-mentioned active entry.

In FIG. 3, an example of an entry added to the LPM table (ie, an active entry) refers to an entry that was previously looked up in the FIB and used for packet forwarding. In addition, as an example of an entry (that is, an inactive entry) deleted from the LPM table in FIG. 3, an entry with a low possibility of use may be used due to a low use frequency in the LPM table.

In other words, the method of detecting the high and low availability can be regarded as a method of determining active and inactive entries separately (in other words, determining an entry to delete and determining an entry to add), which will be described later. .

4 is a block diagram illustrating an apparatus having a three-layer switching function according to an embodiment of the present invention. The apparatus of the present embodiment is a device having a three-layer switching function, and examples thereof include a three-layer switch, a router, and an ONU in an E-PON, but are not necessarily limited thereto.

Referring to FIG. 4, the apparatus of the present embodiment includes an FIB 400, an LPM table 410, a manager 420, and a forwarding unit 430.

The apparatus shown in FIG. 4 may also correspond to components of the apparatus in the implementation illustrated in FIGS. 2A and 2B. For example, the management unit 420 and the forwarding unit 430 correspond to the switching chips 215 and 265, the LPM table 410 corresponds to the second reservoir 270, and the FIB 400 corresponds to the CPU 205. 255) corresponds to a database stored / managed by the first repositories 210 and 260. However, it will be understood by those skilled in the art that the apparatus illustrated in FIG. 4 is not limited to the structure of the apparatus illustrated in FIGS. 2A and 2B.

The apparatus of this embodiment will be described with reference to FIG. 4 as follows.

In the FIB 400, a valid route is stored as an entry as described in FIG.

Among the entries of the FIB 400, highly available entries are stored in the LPM table 410 through the control of the management unit 420. Preferably, the management unit 420 selects an entry of the FIB 400 that is not in the LPM table 410 but is highly available, and inserts the entry into the LPM table 410. In addition, the manager 420 selects and deletes an entry having low availability from among the entries in the LPM table 410. More specific and effective insertion / deletion methods will be described later.

The forwarding unit 430 searches for an entry matching the destination address of the received packet on the LPM algorithm on the LPM table 410 and forwards it to the next hop. Hereinafter, the present invention will be described for convenience by referring to the expression "matching on the LPM algorithm" as "matching".

Preferably, if there is no entry matching the destination address on the LPM table 410, the forwarding unit 430 searches for the FIB 400 and forwards the received packet to the next hop. This step-by-step forwarding concept will be described in detail with reference to FIGS. 5A and 5B.

5A and 5B are block diagrams illustrating an LPM table 410 used in accordance with an embodiment of the present invention. After the LPM table 410 is first searched, an additional FIB 400 may be additionally searched if necessary. A diagram for describing a process of performing forwarding. Specifically, FIG. 5A is a diagram for explaining a step-by-step search method according to a first embodiment of the present invention, and FIG. 5B is a diagram for explaining a step-by-step search method according to a second embodiment of the present invention.

5A and 5B show only the information of the network address and the next hop as an entry in the LPM table, but it is understood by those skilled in the art that various fields may additionally be added to each entry. Can be.

In the default entry of the LPM table shown in FIG. 5A, the forwarding unit 430 is set to refer to the FIB 400 for information on the next hop. Therefore, when the forwarding unit 430 selects a default entry because there is no entry in the LPM table matching the destination address of the received packet, the forwarding unit 430 searches the FIB 400 to determine the next hop of the received packet, and then performs forwarding. Done.

There is no default entry in the LPM table shown in FIG. 5B unlike FIG. 5A. Under the environment of the LPM table shown in FIG. 5B, if there is no entry of the LPM table matching the destination address of the received packet, the forwarding unit 430 automatically searches the FIB 400 to determine the next hop of the received packet. After that, forwarding is performed.

5A and 5B and the LPM table 410 and the forwarding unit 430 for performing the above-described operation (ie, a device to which the stepwise search method according to the first and second embodiments is applied). Since the switching is performed by using the FIB 400 in addition to the LPM table 410, it is possible to perform more effective switching than the conventional method of performing only hardware-based switching, and only software-based switching. Compared to the conventional method of performing, the effect that the overall faster switching can be performed.

Hereinafter, a method of managing the LPM table 410 by the management unit 420 will be described so that an entry highly available among the entries of the FIB 400 exists in the LPM table 410.

The management method of the management unit 420 according to the first embodiment of the present invention is not in the LPM table 410, and among the entries in the FIB 400, a method of inserting a highly usable entry into the LPM table 410. to be. Here, one example of a method for determining an entry with high availability is a method of determining an entry to be inserted into the LPM table 410 by monitoring the frequency of use of the entry retrieved by the FIB 400 and used for forwarding. For example, the manager 420 inserts an entry used for packet forwarding through the FIB 400 into the LPM table 410.

The management method of the management unit 420 according to the second embodiment of the present invention is a method of deleting entries with low availability among the entries of the LPM table 410. Here, as an example of a method for determining an entry with low availability, the management unit 420 monitors the frequency of use of each entry in the LPM table 410 to determine an entry to delete. To this end, preferably, the management unit 420 manages a field (hereinafter referred to as a use field) for recording a use history of each entry of the LPM table 410 in the LPM table 410 and manages it based on the use field. Detect the high or low frequency of availability. A method of detecting the high or low frequency of use based on the use field will be described with reference to FIG. 6.

6 is a diagram illustrating a process of periodically monitoring a usage field according to an embodiment of the present invention.

Referring to FIG. 6, the manager 420 performs an initialization process on the LPM table 410 at a time point t0. As a result of the initialization process, an entry in the LPM table 410 is set, and the hit bit is set to zero. Here, the hit bit is one embodiment of the use field described above.

Referring to FIG. 6, two received packets are input to the forwarding unit 430 between t0 and t1, and the forwarding unit 430 of the LPM table 410 matches the destination addresses of the two received packets. You can see that the entries 10.1.xx and 10.1.1.x were searched and forwarded to the next hop. Here, the management unit 420 sets the value of the hit bit of the entry to 1 to record that entries of 10.1.x.x and 10.1.1.x are used.

At the time t1, the management unit 420 monitors the LPM table 410, obtains the usage history (that is, the value of the hit bit) of each entry, and then sets the value of the hit bit of each entry to 0 ( clearing).

At the time t2 and t3, the management unit 420 similarly performs the monitoring and clearing work.

As described above, the manager 420 may periodically monitor the hit bits of the LPM table 410 to detect availability or frequency of use.

A first preferred embodiment for detecting an entry with low usability based on such periodic monitoring results is a method for determining an entry whose number of consecutive non-uses is greater than or equal to a predetermined threshold as a result of one to several monitoring as an entry with low usability.

A second preferred embodiment of detecting an entry with low availability based on such periodic monitoring results is a method of probabilistically calculating availability based on one to several monitoring results. In one example, the management unit 420 determines that an entry whose cumulative non-use count is greater than or equal to a predetermined number of times of monitoring as an entry having a low possibility of use.

The entry detected as having low availability in this process is deleted by the management unit 420 from the LPM table 410. For example, in FIG. 6, an entry of 10.x.x.x may be deleted from the LPM table 410.

FIG. 7 is a diagram illustrating an LPM table in which a hit bit is recorded and used as a cache according to an embodiment of the present invention.

The LPM table illustrated in FIG. 7 is recorded in the default entry so as to refer to the FIB and includes only an entry having a hit bit of 1. FIG. That is, in the LPM table, an entry that is not used and the hit bit becomes 0 is deleted, and an entry of the FIB used for switching with reference to the FIB is added to the LPM table so that the hit bit is written as 1.

Hereinafter, for more efficient switching, a method of grouping and deleting related entries from the LPM table 410 according to an embodiment of the present invention will be described.

First, a situation that may occur when an entry is deleted from the LPM table 410 by one unit will be described. As an example of this situation, the LPM table 410 has entries 1 = {network address: 10.1.xx-> next hop: B}, entry 2 = {network address: 10.1.1.x-> next hop: C} When included as an entry, there is a situation in which the management unit 420 determines entry 2 as an entry with low availability and deletes it.

When the packet having the destination address of 10.1.1.2 is received after the entry 2 is deleted, the forwarding unit 430 searches the LPM table 410 and confirms that the entry matching the destination address and the LPM from the viewpoint is entry 1. In other words, it passes to the hop corresponding to B. In other words, due to deletion of entry 2, the matching entry becomes L1 in terms of LPM, which causes undesirable packet forwarding. This is because the preferred next hop to which the received packet should be forwarded is the hop corresponding to C of entry2.

Alternatively, if entry 1 is also deleted, the forwarding unit 430 refers to the FIB 400 according to the embodiment of the present invention described with reference to FIGS. 5A and 5B, and as a result, the received packet is a preferred next hop. It can be seen that it can be passed in C.

Thus, to take advantage of this feature, the manager 420 preferably deletes the entries determined to be low in availability together with the associated entries. Here, the associated entry refers to an entry having a relationship between upper and lower on the LPM tree. Preferably, the management unit 420 deletes the entries determined to be low in availability, up to the entry of a predetermined step in the LPM tree rather than the entries. A detailed description thereof will be described with reference to FIG. 8.

8 illustrates an LPM tree for explaining a bundle processing (delete / insert) process according to an embodiment of the present invention.

In the LPM tree illustrated in FIG. 8, there are 10.xxx entry nodes, 10.1.xx entry nodes, etc., which correspond to 10.xxx on the FIB 400 and the LPM table 410. It corresponds to an entry, an entry corresponding to 10.1.xx, and the like.

For example, when 10.1.xx in FIG. 8 is unlikely to be used, the management unit 420 binds 10.1.xx to its upper entry 10.xxx in the LPM table 410 according to the above-described bundle deletion method of the present invention. Delete it.

As another example, when 10.1.1.x in FIG. 8 is unlikely to be used, the management unit 420 bundles 10.1.1.x with its upper entry 10.1.xx according to the above-described bundle deletion method of the present invention, and thus the LPM table. In operation 410, the management unit 420 deletes 10.1.1.x from the LPM table 410 by binding all entries up to the highest entry, that is, 10.1.xx and 10.xxx.

Here, the larger the number of entries that are bundled and deleted, the more the storage space of the LPM table 410 can be secured, while the more the packet forwarding through the FIB 400, the more the switching speed is reduced overall. It can be seen that. Thus, preferably, this property is taken into account to determine the number of entries that are bundled and deleted.

Hereinafter, for more efficient switching, a method (bundle insertion method) of inserting relevant entries into the LPM table 410 according to an embodiment of the present invention will be described.

First, a situation that may occur when an entry is inserted into the LPM table 410 as a unit will be described. An example of this situation is that entry1 = {network address: 10.1.xx-> next hop: B}, entry2 = {network address: 10.1.1.x-> next hop: C} is only present in FIB 400 A situation in which the management unit 420 inserts the entry 1 into the LPM table 410 after detecting that the entry 1 is highly available.

When only a packet having a destination address of 10.1.1.2 is received after only entry 1 is inserted, the forwarding unit 430 searches the LPM table 410 to confirm that the entry matching the destination address and LPM is Entry 1 In other words, it passes to the hop corresponding to B. That is, since entry 2 is not in the LPM table 410, the matching entry becomes L1 in terms of LPM, which causes undesirable packet forwarding. This is because the preferred next hop to which the received packet should be forwarded is the hop corresponding to C of entry2.

Alternatively, if entry 2 is also inserted into the LPM table 410, it can be seen that the received packet can of course be forwarded to C, which is the desired next hop.

Thus, to take advantage of this feature, the manager 420 preferably inserts the entries of the FIB 400 determined to be highly available together with the associated entries and inserts them into the LPM table 410. Here, the associated entry refers to an entry having a relationship between upper and lower on the LPM tree.

Preferably, the management unit 420 bundles an entry determined to be highly available to the LPM table 410 by binding an entry below a predetermined step on the LPM tree rather than the entry. For example, in FIG. 8, if it is determined that an entry of 10.x.x.x is highly available, it is bundled with 10.1.x.x or with 10.1.x.x and 10.1.1.x and inserted into the LPM table 410.

On the other hand, since the number of entry bundles to be inserted has the same characteristics (trade-off relationship between switching speed and storage space) as the number of entry bundles to be deleted, a separate description thereof will be omitted.

Hereinafter, a switching method using a representative entry (hereinafter, referred to as a representative entry method) according to an embodiment of the present invention will be described.

Preferably, in performing the above-described bundle deletion method of the present invention, the management unit 420 leaves only the representative entry among the entries to be deleted according to the method of using the representative entry according to the first embodiment of the present invention, and the FIB 400 Packet forwarding is performed.

9A and 9B are diagrams for describing a concept of performing packet forwarding using an LPM table to which a representative entry is applied according to the representative entry using method according to the first embodiment of the present invention.

9A shows the state of the LPM table 410 before application of the representative entry. In the state of the LPM table 410 illustrated in FIG. 9A, when the manager 420 determines that an entry of 10.1.xx is low in availability, the manager 420 is executed when the above-described method for deleting a bundle of the present invention is performed. ) Deletes 10.1.xx by binding its parent entry 10.xxx. However, in the method of using the representative entry according to the first embodiment of the present invention, the top hop entry to be deleted is left on the LPM table 410, but the next hop field of the top entry is performed so that switching through the FIB 400 is performed. To record. As a result, the state of the LPM table 400 is updated to the state of the LPM table 400 of FIG. 9B.

As a result, the packet which is the destination address of 10.1.2.2 and the packet which is the destination address of 10.2.1.1 are forwarded with reference to the FIB 400.

Meanwhile, in the representative entry usage method according to the second embodiment of the present invention, the management unit 420 bundles at least two entries among the entries of the FIB 400 that are not present in the LPM table 410, and binds the two entries that are bundled together. Representative entries (hereinafter referred to as representative entries) are inserted into the LPM table 410, and the representative entries are recorded so that packet forwarding through the FIB 400 is performed. The recording result is similar to that of the LPM table 400 of FIG. 9B. However, unlike the first embodiment described above, the LPM table 400 does not have 10.1.x.x and 10.x.x.x before the representative entry is applied, unlike FIG. 9A.

In performing the representative entry usage method according to the first and second embodiments of the present invention, there is no limitation on the usage method of the default entry. For example, in the representative entry usage method according to the first and second embodiments of the present invention, the general default entry usage method for setting the next hop corresponding to the default entry to D, the default entry usage method described in FIGS. 5A and 5B, etc. It is applicable.

Such a method and apparatus of the present invention have been described with reference to the embodiments shown in the drawings for clarity, but these are merely exemplary, and various modifications and equivalent other embodiments are possible to those skilled in the art. Will understand. Therefore, the true technical protection scope of the present invention will be defined by the appended claims.

According to the present invention, three-layer switching can be efficiently performed, resulting in the following effects. First, in a device with a limited hardware LPM table, it is possible to support a large size of FIB to enable a more efficient and economical network configuration. Second, the high-performance switching device can be implemented at low cost because it can be freed from the constraint of having a large hardware LPM table to support a large FIB.

Claims (14)

In a forwarding table for three-layer switching, And at least one entry having recorded field values for referring to a separate forwarding table. The entry of claim 1, wherein the entry in which the field value to be referred to is recorded. A forwarding table for layer 3 switching, comprising a default entry. The entry of claim 2, wherein the entry in which the field value to be referred to is recorded. And at least one of the remaining entries other than the default entry. The entry of claim 1, wherein the entry in which the field value to be referred to is recorded. And at least one of the remaining entries except for the default entry. The entry of claim 1, wherein the entry in which the field value to be referred to is recorded. A forwarding table for three-layer switching, comprising an entry (hereinafter, representative entry) representing upper and lower entry nodes consecutive in a predetermined step on the LPM tree. The network address field of claim 5, wherein And a network address corresponding to the highest entry node of the consecutive upper and lower entry nodes is recorded. The method of any one of claims 1 to 6, wherein the separate forwarding table is A forwarding table for three-layer switching, comprising a forwarding information base (FIB). A forwarding table used for forwarding a packet received from a network to the next hop to reach its destination. Network address field; Next hop field; And A field indicating whether the network address field is used or not; At least one of the next hop fields has a value for referring to external forward information. The method of claim 8, wherein the field indicating whether or not to use A forwarding table, characterized in that it is set to indicate the used state. 10. The forwarding table of claim 8 or 9, wherein a network address field corresponding to a next hop field set to reference the external forward information is set in a default form. The forwarding table of claim 10, wherein the network address field corresponding to the next hop field set to refer to the external forward information is set in the form of an IP address. 12. The forwarding table of claim 11, wherein the external forward information is a forward information base (FIB). 13. The forwarding table of claim 12, wherein the forwarding table is a Longest Prefix Matching (LPM) table. The forwarding table of claim 13, wherein the field indicating whether the network is used is set to indicate a state in which the network address is used through a frequency used for forwarding.

Images