WO2003017584A2 - Signalled structure of connection sequences - Google Patents

Abstract

In a connection network, in which connection nodes are connected by connection runs forming connection tunnels, sequences of several connection tunnels may be structured to form transit tunnels for data transport by sequences of virtual links using individually signalled information.

Description

description

_S ignalisierter building junction sequences

_A nmeldungsgegenstand relates to a method Einrich ^¬ th of connections in a connection network are connected in the connecting nodes via links.

_On the basis of a physical communication network consisting of Ml .. m physical links and Nl..Nn physical nodes, several virtual communication networks (VN) are to be set up. Each VN consists of a number of connections (such as m (bidirectional) ATM (Asynchronous transfer mode) -SVCs (switched virtual circuit, virtual dial-up connection) or 2 * m unidirectional MPLS (multi protocol label switching) -LSPs), i.e. m or 2 * m virtual links, which connect a total of n <= N physical nodes. Each unidirectional MPLS-LSP is accompanied by its unidirectional reverse virtual link.

All telecommunication networks are physical "partial mesh * networks". A full mesh “fill mesh * would be nonsensical. A certain number of connections or tunnels can be established on a physical network, which as a whole each result in a so-called virtual network. This virtual network comprises a total of n nodes. With full meshing of the virtual network one would need n * (n-l) / 2 bidirectional connections or n * (n-l) unidirectional connections. This “virtual fill mesh” is still too complex (too expensive).

For comparison, a few “partial mesh” networks: A minimal virtual partial mesh only requires n unidirectional connections (= ring) or n-1 bidirectional connections (= tree). A somewhat more complex "Karo" network only requires 2 * n bidirectional connections or 4 * n unidirectional connections. Examples a) Virtual Private Network (VPN) .Genauer: the Connection ^¬ network between the n locations a private company, wel ^¬ ches the "public network bridges *.

b) Virtual Service Provider Network (VSPN). More precisely: The connection network between the n switches / routers of a service provider, which bridges the network of another service provider. This connection network is also called the Carrier's Carrier Network.

c) VPN based on a VSPN network

d) MPOA overlay model of the ATM Forum: -Applications: In the Multi-Protocol-over-ATM (MPOA) model of the ATM Forum, a lot of ATM-SVC connections are used to establish an ATM network as a kind of transit network for the Internet to use. The ATM switches are also connected to Internet routers. IP packets are transported in the form of ATM cells through the ATM network in order to then forward them as IP packets in an IP network. Previous solutions: a) fill mesh ie n * (nl) / 2 SVCs, or b) SVC management, which can set up a maximum of m SVCs. If a SVC is required that does not yet exist, it will be set up if the maximum m has not yet been reached. Otherwise, either the transfer request rejected or it is broken down to create an existing SVC to "place ^* - about an SVC has the relative little Vehrkehrslast.

In the case of "partial mesh *", the user data to be transported must pass through a certain sequence of connection tunnels in order to arrive at a predetermined destination node.

The subject of the application is based on the problem of producing a quasi full mesh, even if the full For example, for reasons of savings, it is far below the ^minimum .

The problem is solved by the features of the characterizing part of claim 1.

When Subject Matter therefore more connection ^¬ will stretch to sequences of connecting tunnels built and used as a transit tunnel, in which case consequences of virtuel ^¬ len links can be run through the data transport.

The subject of the application advantageously saves a lot of tunnel connections.

Advantageous further developments of the subject of the application are specified in the subclaims.

Such a sequence can in turn be viewed as a tunnel.

This can be further increased recursively. This leads to the concept of the hierarchical tunnel (from a certain hierarchical level).

A hierarchical tunnel is therefore a built-up sequence of component tunnels = element tunnels. "Structure * or" structure sequence * should mean: Tunnel link settings are made (label switching entries) in such a way that the user data stream can change the individual tunnels just as quickly and easily as the lines within a tunnel.

The subject of the application is explained in more detail below as an exemplary embodiment to the extent necessary for understanding with reference to figures. Show:

1 shows an abstracting representation of a switching network in which the invention can be implemented and 2 special features according to the invention in the formation of sequences of connecting lines.

In Figure 1, the thin lines represent physical connection ^¬ stretch and their nearest physical router of a _V determination network. The solid arrows denote one-way links (links) and the broken arrows hops of the hierarchical multi-point-to-point LSP (LSP = label switched path).

The individual connections / tunnels can be of various technologies:

ISDN connection, ATM-SVC, frame relay SVC, MPLS-LSPs, Internet tunnels according to the IETF standards such as IPSEC, L2TP, PPP, GRE, IP in IP.

A connection / tunnel sequence established by signaling can be technology heterogeneous

In a preferred embodiment, a connection / tunnel sequence established by signaling is independent of the transmission standard (s) used.

The following connection / tunnel sequences can be set up: point-to-point sequence, point-to-multipoint sequence, multipoint-to-point sequence.

Point-to-multipoint: The user data through the tree-like tunnel sequence set up flow from one transmitter tunnel exit point to several receiver tunnel end points.

Multipoint-to-point: The user data through the tree-like tunnel sequence set up flow from several transmitter tunnel exit points to a receiver tunnel end point.

A tree-like connection / tunnel sequence can either be set up from the tree root to the leaves or starting with the leaves to the root (= reverse path setup). Both variants come for both point-to-multipoint and for ultipoint-to-point tunnel ^¬ sequences into consideration.

Subject of the application is in uni-directionality and bidi rektionalität of connections / tunnels alike Execute ^¬ bar.

Recursivity:

A hierarchical tunnel of the lowest (= first hierarchical level) consists of a linear sequence of hops along physical lines. A tunnel of a higher hierarchical level consists of a linear sequence of element tunnels of a lower hierarchical level.

The element tunnels of a tree-like tunnel sequence can e.g. Tunnels of the 1st hierarchical level (VPN over the physical network of a service provider.)

The element tunnels of a tree-like tunnel sequence could also be of a higher hierarchical level (VPN over a virtual service provider network).

Each element tunnel includes identifying and characterizing data of the "globally significant *" and "locally significant *" types.

The globally significant data may include:

Ru = router address of the upstream tunnel end point (where the user data go into the tunnel),

Rd = router address of the downstream tunnel end point (where the user data leaves the tunnel),

Color = abstract summary of all characterizing properties such as bandwidth, QoS (Quality of Service), purpose like "for voice * or" for data *, etc. The (different) color becomes important especially when it is there are several parallel tunnels with identical soot and identical round.

L _S P-ID (Label switched path identifier) (= tunnel identifier)

It consists of the address Ru plus a number assigned by this upstream router.

(We want to get the tunnel sequence to be set up, as it were, from a higher perspective, as only one tunnel is viewed and thus also get an LSP-ID just like each of its element tunnels. This LSP-ID is, according to the invention, the router at the downstream end the tunnel sequence.

The globally significant data from many tunnels (such as all tunnels that are used exclusively for a very specific VPN) may - statically or dynamically - be passed on to a process that may determine certain tunnel sequences from them.

Locally significant data from a tunnel:

These are the label switching entries for the respective tunnel switching. They can be found at the tunnel endpoints using the tunnel ID.

However, the tunnel sequence that is built up (which consists of these element tunnels) also receives such globally and locally significant data.

Label allocation:

The downstream router issues a label. He tells the upstream router "with this label you should send me the user data packets *.

A tunnel sequence can be considered as established if the following is installed = saved: LIB entry on the router of the upstream end of the tunnel sequence: (LIB = Label Information Base)

_{a) S} uc _h-key (comparison value) to LIB-entry will be at the:

FE _C information, i.e. which destination address areas can be reached at the downstream end of the tunnel sequence (FEC = Forwarding Equivalence Class)

b) Wanted:

Label relating to the first tunnel of the tunnel sequence;

Label of the first hop of the first tunnel;

Physical management (interface index) of the first hop of the first tunnel;

LIB entry on the router that receives the data via an incoming tunnel and forwards it via an outgoing tunnel:

Search key (table index value) to find the LIB entry:

Incoming tunnel label

b) Wanted:

Label regarding the outgoing tunnel of the tunnel sequence;

Label of the first hop of the outgoing tunnel;

Physical management (interface index) of the first hop of the outgoing tunnel.

If the element tunnel itself already represents a sequence of even more elementary tunnels, another label (in second position) can be found under what is searched.

The tunnel sequences are set up by signaling.

The construction of a non-branching / non-uniting tunnel sequence - i.e. a linear tunnel sequence - is basically the same as building a tunnel at the very first hierarchical level.

_A tunnel element us a VPN be constructed uniting doing ^¬ nelsequenzen that - no matter what the output node ^¬ beginning - the user data, to a predetermined destination node are transported.

For each node Z (such as Target) of VNs the question arises: "over which series of virtual links should ever ^¬ of the other n-1 node S (such as start) hintransportieren of VNs transfer data to it together, all these sequences form? of virtual links a tree rooted in Z

"Settings" should be made at all nodes that appear in this tree so that the data transport can proceed as desired. The sum of these settings results in a so-called hierarchical multipoint-to-point connection.

A hierarchical multi-point-to-point LSP enables the following:

Any node S sends data intended for Z by giving these two MPLS labels. The top label in the label stack is responsible for label switching at the next physical node of the first virtual link, i.e. it is used in the next physical node to determine the physical continuation link and the continuation label.

The lower label is carried transparently, i.e. unchanged, right through to the end of the first virtual link. There the upper label is removed and the lower label is used to access the following information: from the correct virtual follow-up link (in direction Z) the first physical line and its first label (which functions as the new upper label should greed. Furthermore, a new lower label, which should function as just described at the end of the second virtual link.

In the case of ATM everything applies perfectly analogous, with two possible answer ^¬ th implemented are:

1) The upper label is a VPI (Virtual path identifier), the lower label is a VCI (Virtual Channel identifier). In other words, the virtual links are SVPs. The ^tree-¬-like effects of SVPs form a hierarchical multi ^¬ point-to-point SVC.

2) The upper label is a VCI, the lower label is a VPI. In other words, the virtual links are SVCs. The tree-like sequences of SVCs form a hierarchical multi-point-to-point SVP.

Before the hierarchical multi-point-to-point connections can be established, the basic connections to be "used", that is, the virtual links, must first be known and established.

A unique ID is assigned to each virtual link. It consists of "address of its ingress node + a unique number assigned by it itself". The IDs of two virtual links must be different even if they belong to different VNs.

In the case of MPLS, a suitable virtual link in the form of its ID must also be known for each virtual link. Suitable means: they must both belong to the same VN (only having the same end point nodes should not be sufficient).

This assignment knowledge must in particular have and be able to evaluate the two end nodes of the respective virtual link, but also each node Z (or a VN for this central proxy server), which is to determine the tree of a hierarchical multi-point-to-point connection.

The destination node Z is calculated as the tree of virtual links (or can it from a central server-zBVN be expected ^¬).

_S o as there may be a plurality of parallel virtual links (for example, one for data, a second for speech), so there can be multiple hierarchical multi-point-to-point connections with the same destination node (eg, one for data, a second for speech).

There are numerous ways in which this hierarchical multi-point-to-point connection can be established:

1) In an MPLS network with unidirectional virtual links la): "From Z to n-1 start nodes in a single operation

Starting from Z, a minimal number of setup messages (e.g. as many virtual links as there are adjacent to Z) are sent, and this is provided with source routing information that satisfies the upcoming branches. In the case of MPLS-LDP, these could be "unsolicited" Label_Mapping messages.

Node Z send out an "unsolicited" Label_Mapping message. It contains a suitable TREE ROUTE TLV which corresponds to the branches to be walked on. It also includes, for example, in a specially defined Passed_Virtual_Link-TLV the ID of the virtual link via which the user data should arrive at node Z.

As far as the source routing information is concerned, the hierarchical aspect is added here: The individual ER-HOP TLVs in the source routing information (TREE ROUTE TLV) may always contain the IDs of two inverse virtual links, namely the ID of the uplink (in the direction S) and the ID of the virtual downlinks (towards Z). When calculating the tree with root Z, care should be taken to ensure that all downlin _k s meet the requirements of user data transfer, ie that the correct one is determined from several parallel downlinks, for example. The relevant (inverse) uplink, however, may be any one, or may always be the first listed or one that is labeled "for data".

Assume that a node K-down has received the label mapping message for the establishment of a hierarchical multi-point-to-point TLV and has to evaluate a very specific ER-HOP TLV (if x-links are to be combined in K-down, so are es x ER-HOP TLVs). The ID of the uplink tells him the physical start line including the start label in order to convey the message, appropriately modified, label-switched via the uplink to the node K-up. Before doing this, however, he determines an available label V _Incomng with regard to the entire virtual downlink, which he intends to communicate to the node K-up via label TLV. The node K-down has received a label V outgoing itself via label TLV. Node K-down take from the Passed_Virtual_Link-TLV of the self-received setup message the ID of the virtual link that it is to use to forward the user data in the Z direction. Of course, he knows the physical start line p outgoing including the start label P _o utgoi _ng from this virtual link. It creates a label information base entry (LIB entry) which can be found using V _ιn comιng and which contains:

"Replace V _ιncomιng with V outoing and then push P _outgoιng into the label stack and forward the data packet with both labels via p outgoing".

K-down sends the label mapping on to K-up, where it writes in the Passed_Virtual-Link TLV the ID of the virtual link via which it intends to receive the user data from K-up. If x should unite virtual links in K-down, he determined for each of x "incoming" links analogous one available label V _ιncomng and a LIB entry in wel ^¬ chem is:

"Replace THIS V _ιncOM „ g with V outgoing and then push P _{ou go} i _ng n the label stack and forward the data packet with ^¬ the labels via p outgoing ".

In this case, each of the x different nodes K-up is informed of the ID of the respective virtual link via Passed_Virtual_Link-TLV, via which it is to send the user data to K-down.

lb) From Z to n-1 start nodes in n-1 staggered processes

Starting from Z, n-1 setup messages are transmitted in chronological succession in order to build up exactly one branch to exactly one start node S in accordance with the "ADD PARTY / SETUP" concept of the ATM Forum.

In order to lean more on the MPLS terminology, the setup message is also an unsolicited label mapping message. It contains two types of source routing information

1) An ER-TLV whose ER-HOP TLVs always only contain the IDs of the respective uplinks.

These are used to convey the message across those virtual uplinks whose associated downlinks have already been taken into account in the hierarchical multi-point-to-point connection established so far (for the "ADD PARTY section)

2) An ER-TLV whose ER-HOP TLVs contain the two IDs of the respective uplinks plus downlinks as described above (for the "SETUP" section).

Passed_Virtual_Link-TLV contains the ID of the downlink that is inverse to the very last uplink of the former ER- TLVs, i.e. the ID of the virtual link that is already integrated in the hierarchical multi-point-to-point connection and in which the current extension branch should flow.

The former ER-TLV does not apply to those build-up messages that build a branch up to node Z. The PasD_Virtual_Link-TLV then contains the ID of the downlink via which Z is to receive the user data.

1c) From the n-1 start node to the destination node Z

Starting from Z, messages are sent to the n-1 nodes S, which contain encapsulated "join" messages which, in accordance with the point-to-point-shaped source routing information contained therein, are each intended to set up a branch in the direction Z - as it were, on the principle of the "reverse path setup". All of these "join" messages must contain one and the same identification for the jointly created hierarchical multipoint-to-point connection, for example consisting of a doublet "address of Z + a unique number assigned by Z". This ID must be unique, also given the large number of VNs that exist at the same time. As soon as a "join" message arrives at a node that has already passed an earlier join with the same LSP ID, it may no longer be forwarded in the direction of Z.

The join message contains exactly one ER-TLV (Explicit Route TLV) whose ER-HOP TLVs are the IDs of the virtual links to be passed. The label must be assigned according to independent mode - anything else would be nonsense given the existence of variants la) and lb). That

Each join message is similar to a label request, which makes a label assignment after each HOP and reports exactly one hop back towards node S (by means of a label mapping message) to make the appropriate LIB entry there.

The join message uses the virtual links for their own progress, which are also intended for the user data stream. A Passed_Virtual_Link-TLV may contain the ID of the virtual link (downlinks) that the join message has just used. A node K-down, which has just received it, will be able to determine the ID of an inverse virtual link (uplink) and can use it to send a label mapping message exactly one hop back towards S. This will be equipped with a label TLV in which K-down has entered a label V _{lncomιnς which it} has assigned itself according to the principle of independent modes. For its part, K-down waits for a label mapping message and expects the analog label V outgoing from there. • It also knows the ID of the virtual link leading directly in the direction of Z, and thus its start line p outgoing together with the start label P outgoing based on the first ERHOP TLVs obtained.

K-down forms the LIB entry, which can be accessed via _video and which contains:

"Replace V _λncomιng with V outgoing and then push P _outgoιng into the label stack and forward the data packet with both labels via p ou oing".

2) In an ATM network with bidirectional virtual links The method described in lb) is possible. Instead of ERHOP TLVs, PNNI DTLs are used.

With full meshing one has a VN whereby the number of virtual links is of the order n ²

Quasi-full meshing is only required for the order n many virtual links.

The VN could be both a Virtual Service Provider Network (VSPN) and a Virtual Private Network (VPN). Furthermore: VSPNs could in turn operate VPNs. The VSPN consists of a number of 1st degree virtual links. A VPN operated by him consists of second-degree virtual links. Such a 2nd degree virtual link generally consists of several 1st degree virtual links. The 2nd degree virtual link is a hierarchical 1-point-to-point connection, the structure of which works in exactly the same way as the structure of the hierarchical (n-1) multi-point-to-point connection described above, i.e. n = 2 ie nl = l.

The quasi-full meshing of such a VPN would only be possible through hierarchical multi-point-to-point connections, which connect tree-like virtual links of the 2nd degree to one another in such a way that the user data can be transported to Z, namely totally label-switched.

LIB entries of the type are typical:

"Replace V _ιn commg with V outoing and also push the labels W outgoing and P outgoing into the label stack and forward the data packet with both labels via p ougoing". W outoing is the label for the first virtual link of the 1st degree from the relevant virtual link of the 2nd degree.

Other possible uses are in the field of traffic engineering:

For whatever reason, an administrator has already installed numerous tunnels (= virtual point-to-point links). At a later point in time, instead of establishing new 1st degree tunnels of this kind, he would like to take advantage of the existing ones and establish certain tunnel sequences as a quasi new tunnel.

Claims

claims 1. A method for setting up connections in a connection network in which connection nodes (Nl..Nn) are connected via connection links (Ml..Mm), characterized in that a sequence of several connection links can be set up by means of signaling in response to individual signaling information , 2. The method according to claim 1, characterized in that a connection / tunnel sequence established by signaling is independent of the transmission standard (s) used. 3. The method according to any one of the preceding claims, characterized in that the sequence of a plurality of links is set up between exactly one start connection node and exactly one destination connection node. 4. The method according to claim 1 or 2, characterized in that the sequence of a plurality of connecting links is set up between exactly one starting connection node and a plurality of destination connection nodes. 5. The method according to any one of claims 1 or 2, characterized in that the sequence of a plurality of links is established between a plurality of start connection nodes and exactly one destination connection node. 6. The method according to any one of claims 4 or 5, characterized in that the establishment of a connection / tunnel sequence from which exactly one connection node to the plurality of connection nodes takes place. 7. The method according to any one of claims 4 or 5, characterized in that the establishment of a connection / tunnel sequence from the plurality of connection nodes to the exactly one connection node. 8. The method according to any one of the preceding claims, characterized in that the designation (LSP-ID) for the connection / tunnel sequence is assigned by the network node at the downstream end of the connection / tunnel sequence. 9. The method according to any one of the preceding claims, characterized in that a sequence of links as a unit together with at least one further link forms a new uniform sequence.