Multiprotocol label switching

from Wikipedia, the free encyclopedia
MPLS on the TCP / IP protocol stack
 application  HTTP BGP LDP
 transport  TCP UDP
 Internet   IP 
 -   MPLS 
 Network access  ATM FR Ethernet

Multiprotocol Label Switching ( MPLS ) enables the connection-oriented transmission of data packets in a connectionless network along a previously established (“signaled”) path. This switching process is mainly used by operators of large transport networks who offer voice and data services on the basis of IP (large Internet providers ).

Technical basis

MPLS was introduced in order to be able to use various advantages of connection-oriented switching in otherwise connectionless networks. This includes, on the one hand, enabling faster processing of a parcel through simplified addressing with the help of so-called labels ; this advantage over traditional longest prefix match forwarding has meanwhile been put into perspective by advanced technology ( ASICs ). On the other hand, MPLS enables the network operator to specify defined paths in his network, which is not possible with traditional routing protocols such as Open Shortest Path First (OSPF) or IS-IS . A distinction is made between the following types of data transfer with or without a preset connection:

Transmission without a fixed connection
If data can be sent spontaneously from a terminal to a recipient, and every intermediate network node (usually router ) knows independently how to forward the data, this is called connectionless data transmission. In a connectionless network, data can in principle arrive at any time and in any amount in the switching system, which is why the availability of the resources required for the transport of certain data cannot be guaranteed. This network has a stochastic and rather random behavior.
Fixed link transmission
If a path through the network to the recipient has to be signaled before a terminal device can send data, this is referred to as connection-oriented data transmission. In this case, the network nodes (usually switches ) are provided with the necessary connection information in order to be able to correctly forward the sent data ( label switching ). The network shows a deterministic and controllable behavior. Resources in the switching systems can be reserved during the signaling phase.

history

In the mid-1990s, the proportion of voice communication ( telephony ) clearly outweighed the proportion of data communication in large-area communication networks ( WANs ) . Due to the differences between connection-oriented (→ telephone calls) and connectionless (→ data packets on the Internet ) transmission described above, telecommunications companies operated separate networks for data and voice transmission, which resulted in considerable costs. A network-wide quality of service (QoS, Quality of Service ) did not exist. Existing voice networks offered these quality of service for voice services, but the bandwidths required for data transmission were either not available or extremely expensive.

The introduction of ATM solved this problem in many areas. Voice and data could now be transmitted over a common infrastructure. However, the ATM transport network did not provide any IP routing functionalities for IP-based data transmission (Internet). This continued to happen in routers .

By using ATM, however, the routing systems were given the opportunity to use much higher data transmission rates. The signaling of connection paths is left to the ATM network, while the IP routers transmit their IP data packets without connection, i.e. stochastically. On the other hand, there is no network-wide quality of service to integrate data and voice using the high bandwidths . This resulted in so-called overlay architectures in which the IP layer uses the underlying ATM transport layer, but both act independently of one another. Examples of this overlay approach are IP over ATM [ RFC 2225 ] and Multiprotocol over ATM (MPoA).

The available router systems reached their capacity limits due to the newly available high bandwidths. In addition, the breaking down and assembly of IP packets (up to 1536 bytes or more) in ATM cells (53 bytes) was a difficult-to-overcome limit for speeds above 622 Mbit / s. Due to the high number of point-to-point -Connections between the routers, especially in fully meshed networks, the use of traditional topology / routing protocols ( IGPs ) such as OSPF , RIP or IS-IS leads to considerable additional signaling traffic ("n-square problem": there are points between points complete meshing of edges). As a result, routers collapse or become permanent bottlenecks in the network. The transmission of different services (voice, data, video) via a standardized and simplified platform does not exist.

MPLS offers solutions for the problem points mentioned above.

Basic idea

Since the late 1990s, MPLS has offered the possibility of relieving overloaded routing systems and thus better utilizing the available bandwidths of the wide area transmission lines.

The idea is to no longer forward data packets from one router to the next router (hop-by-hop), and to make the decision again for the cheapest route in each router (complete IP lookup in the so-called forwarding table) , but to send them to an entry point ( ingress router) on a pre-signaled data path and only use the conventional hop-by-hop forwarding of IP at the starting point ( egress router). Ideally, ingress and egress routers are located at the boundaries of a network. This procedure relieves a large part of the router: on all MPLS-capable intermediate stations, so-called label-switched routers (LSR), only the labels upstream of the MPLS packets are evaluated. This takes place directly above the data link layer (Layer 2) and can be done very easily in the appropriate hardware at high speed - in contrast, forwarding decisions in traditional IP routing require the considerably more complex Longest Prefix Match .

MPLS offers connection-oriented traffic behavior like ATM for data packets. The paths are set up (signaled) once before the packet is forwarded and are then available. In addition, with the help of additional protocols or protocol extensions, such as CR-LDP or RSVP-TE, resources can be reserved on the routers or the route selection can be specifically influenced. To a certain extent, this allows QoS to be implemented network-wide for the combined transmission of voice, data and video.

Nevertheless, MPLS cannot reserve bandwidths even with RSVP, as ATM allows. It is possible to approximate a certain deterministic traffic behavior, but IP routing / forwarding is stochastic in its behavior, even when using MPLS.

The initial speed advantage of MPLS in data forwarding is no longer relevant nowadays, as modern routing systems have implemented IP forwarding in hardware throughout.

functionality

The use of MPLS in IP networks requires a functioning logical and physical IP-based network infrastructure (MPLS-capable router). MPLS acts primarily within the limits of what is known as an autonomous system (AS). In addition, the use of an Interior Gateway Protocol ( IGP ) such as B. OSPF or IS-IS, makes sense. Theoretically possible, but not very practical, would also be the use of static routes in combination with IBGP.

Structure of the MPLS paths

After it has been ensured that the routers of an autonomous system (AS) can all "see" each other (this is ensured , for example, by OSPF or IS-IS), the MPLS paths (paths) are now switched between the individual routers. These paths are called Label Switched Path (LSP). The start node of an LSP is known as the ingress router , the end point as the egress router . These start and end nodes are typically located at the entry and exit points of an AS (AS Boundary Router) .

The LSPs can be switched completely manually, semi-automatically or fully automatically. The manual variant requires the configuration of every router that an LSP runs through. This method is inefficient for autonomous systems on the order of several dozen routers. The semi-automatic variant only requires the manual configuration of parts of the LSP, for example the route via the first three routers. The rest of the pathfinding for the LSP is left to the IGP. The fully automatic variant completely relies on the IGP when determining the path for an LSP. Thus, there is no advantage in terms of path optimization. However, data is now forwarded in the routers on layer 2 (label swapping, i.e. exchanging / changing labels) instead of on layer 3.

Routing IP packets

As soon as an IP packet enters an MPLS network, it is provided with an additional MPLS header ( see below ) on the ingress router . If you consider the ISO layer information (see also ISO / OSI reference model ) of a data packet, this header is between the layer 3 information (network layer header) and the layer 2 information (link layer Header) inserted. This process of inserting is called a push operation . If the label of an LSP is removed by a router, this is called a pop operation . The exchange of the label by a router on the path of an LSP is called a swap operation . This requires that Label 2 and Label 3 each store their own connection information (bandwidth, latencies and destination) in order to be able to forward data via MPLS without errors.

Penultimate hop popping

Penultimate Hop Popping (PHP) describes the fact that an MPLS label (the outer label in the case of stacked LSPs) is already removed in the penultimate router of an LSP. This so-called PHP router knows the path to the egress router, due to the IGP, and forwards the data packet to it in the normal way. This saves the POP operation in the egress router, it only has to forward the unpacked packet based on the routing information.

Development of MPLS

The advantage of MPLS only becomes apparent when additional services based on MPLS technology are used. Such - meanwhile largely standardized - services are currently:

Traffic engineering
Is the targeted control of the route selection for the data traffic of a network. This application enables a network operator, for example, to offer its customers particularly broadband and low-delay data paths in a targeted manner. For example, RSVP- TE can be used to provide resources for optimized routes through a network .
Layer 2 VPN
Virtual private networks (VPNs) on OSI layer 2, with point-to-point connections. These enable ATM connections (VPI / VCI), Ethernet VLANs or Frame Relay paths (VCs = Virtual Circuits) of different networks to be connected to one another directly via an IP-MPLS network. A connection on layer 2 of the OSI model is handed over to the customer at the transfer point. An application example would be an Internet service provider who offers DSL access in Germany but does not have its own nationwide infrastructure to connect its ATM-based DSLAMs (Digital Subscriber Line Access Multiplexer) to a central BRAS . To do this, he then uses another transport provider with an IP-MPLS infrastructure, which transparently routes the ATM VPIs / VCIs to the central BRAS location. The DSLAMs and the BRAS system are provided with an ATM interface by the transport provider, although their infrastructure is purely IP-based. This is also referred to as pseudo-wire emulation (PWE3 Circuits). Lines / paths are thus emulated. In this case, the ingress / egress routers are referred to as label edge routers (LER), the routers on the path of the LSP as label switch routers (LSR). In practice, the LSPs are automatically signaled between the LERs using a separate protocol (e.g. LDP or L2TPv3). It is also possible to configure the LSPs manually. A Layer 2 VPN is like a virtual ATM / Frame Relay / Ethernet switch with point-to-point connections.
Virtual Private LAN Service (VPLS)
This is a variant of Layer 2 VPNs with a focus on Ethernet-based infrastructures, i.e. point-to-multipoint connections, which takes into account the broadcast behavior of Ethernet. LDP is predominantly used as the signaling protocol, but BGP is also used. At the transfer point, the customer is provided with a bridged LAN port. A VPLS instance is like a virtual LAN switch.
Layer 3 VPNs
Virtual private networks (VPNs) on the OSI layer 3. These make it possible to map complete, routed network infrastructures of customers transparently via a provider MPLS transport network. A connection on layer 3 of the OSI model is handed over to the customer at the transfer point, i.e. a routed connection with a static route or an IGP. Further details can be found, for example, in RFC 4364 . In practice, the LSPs are signaled using LDP. A Layer 3 VPN looks like a virtual IP router (which should not be confused with the proprietary "virtual router" concepts of some manufacturers).
G-MPLS (Generalized MPLS)
It expands the scope of MPLS to include optical transmission infrastructure. This approach should include the automatic signaling of optical paths (for example, individual wavelengths of a WDM / DWDM interface, SDH paths or a complete interface) when setting up an LSP. The signaling of the topology extends its scope away from the IP transport layer within an AS to the underlying infrastructure transport layer. Standardization approaches for the architecture, the functional model and requirements for this can be found under the search terms ASON / ASTN (Automatic Switched Optical Network / Automatic Switched Transport Network).

Basically, MPLS technology leads the independent packet switching (connectionless) back to the line switching by LSPs (connection-oriented). This limits some of the advantages of IP-based communication from anyone to everyone ( any -to- any ) , with all its flexibility and good scalability, by the strengths of connection-oriented communication (complexity, n-square problem, etc.).

MPLS solutions are now being implemented in almost all large end customer WANs. However, in order to be able to continue to meet the steadily increasing demand for higher bandwidths, the conventional WANs have been further developed into hybrid WANs, in which the advantages of MPLS technology are combined with those of VPNs.

Structure of MPLS packages

With MPLS there are basically two different options for labeling a package. One, for example with IP, provides a so-called MPLS shim header , which is inserted between the layer 2 header and the layer 3 header. However, this header is usually called the MPLS label stack (entry) . In the case of connection-oriented networks, on the other hand, such as ATM or Frame Relay, the label can be added to the Layer 2 header; there is then no separate MPLS label stack entry.

MPLS label stack entry

The MPLS label stack entry is not actually a header; the word shim expresses how short it is. It has a length of 4 bytes (32 bits), so it generates little overhead and can also be processed very quickly. The heart of the MPLS label stack entry is the MPLS label . In particular, the label determines which path (LSP, Label Switched Path ) should be used to route the packet through the MPLS network.

MPLS Label Stack Entry (MPLS Shim Header)
0 1 2 3 4th 5 6th 7th 8th 9 10 11 12 13 14th 15th 16 17th 18th 19th 20th 21st 22nd 23 24 25th 26th 27 28 29 30th 31
Label TC S. TTL

With the 32 bits of the MPLS label stack entry four additional pieces of information are conveyed:

  • Label (MPLS label; 20 bit): identification information of an LSP (comparable to a telephone number). It is important to understand that this label only has local validity, i.e. it is only used between two routers on the path of an LSP and not on the entire path from the ingress to the egress or PHP router.
  • TC (Traffic Class; 3 Bit): Used to transmit Differentiated Services information.
  • S (Bottom of Stack; 1 bit): Defines whether the LSP is a nested LSP, i.e. whether another LSP is transported in the LSP. The flag indicates whether there are further MPLS labels to follow or whether this MPLS label stack entry represents the last label of the label stack.
  • TTL ( Time to Live ; 8 Bit): Defines how many MPLS routers the packet can still pass through (limit: 255 routers)

MPLS label stack

Usually exactly one label is assigned to each package. However, if you want to nest several LSPs within one another, you can assign several labels to an MPLS packet. These are then summarized in the so-called label stack :

MPLS label stack

The use of the bottom of stack flag is clearly visible here . The evaluation takes place from left to right, after the "Bottom of Stack" the Layer 3 header follows directly.

Embedding the MPLS label stack

Depending on whether a nested or a simple LSP is present, an MPLS label stack consisting of one or many MPLS label stack entries is inserted.

 Layer 2 header (e.g. Ethernet)   MPLS Label Stack / MPLS Label Stack Entry   Layer 3 header (e.g. IP)   Layer 4 header (e.g. TCP)   Payload 

Norms and standards

See also

literature

  • Nam-Kee Tan: MPLS for Metropolitan Area Networks. Auerbach Publications, CRC Press, 2004, ISBN 0-8493-2212-X .
  • Uyless D. Black: MPLS and label switching networks. Prentice Hall PTR, 2001, ISBN 0-13-035819-3

Web links

Individual evidence

  1. ↑ MPLS Multiprotocol Label Switching
  2. ↑ Hybrid WAN - combination of MPLS and Internet. Retrieved September 17, 2018 .