L2VPN Working Group Internet Draft J. Cho Document: draft-jaihyung-l2vpn-lse-00.txt ETRI Expires: April 2004 October 2003 A Label Switching Technique for Ethernet Frame Transmission Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026 [i]. This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026 except that the right to produce derivative works is not granted. This document is an Internet-Draft and is NOT offered in accordance with Section 10 of RFC2026, and the author does not provide the IETF with any rights other than to publish as an Internet-Draft Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet-Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. Abstract LSE (Label Switched Ethernet) is a label switching technique for transmission of Ethernet frame. The method offers not only L2VPN connectivity but it also provides features for dynamic connection establishment and resource reservation. LSE does not require additional information fields other than 802.3 standard header because it uses destination address and source address as labels. As a result, compatibility with legacy Ethernet is maintained and it provides label switching service from terminal to terminal. Equipments that support LSE are connected via Ethernet tunnel merging Cho Expires - April 2004 [Page 1] LSE October 2003 tree or TLSP (Tunnel LSP) tree across legacy network equipment. User terminals establish connections using modified ARP protocol. Use of merging tree structure helps reducing signaling overhead as well as improving scalability of LSE network. LSE also provides unified mechanism for unicasting, multicasting and wireless terminal mobility. Conventions used in this document In examples, "C:" and "S:" indicate lines sent by the client and server respectively. The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC-2119 [ii]. Table of Contents 1. Overview.......................................................2 2. Classification of Network Equipments...........................3 3. Label Switched Ethernet Frame Format...........................4 4. LSE Tunnel Establishment.......................................4 5. Signaling & Connection Setup...................................6 6. Label Switched Data Transmission .............................8 7. Data Transmission in All LSE Network...........................9 8. Support for Multicasting and Wireless LAN Mobility............12 9. Structuring LSE Network.......................................14 10. Conclusion...................................................15 11. Security Considerations......................................15 12. References...................................................15 Author's Addresses...............................................16 1. Overview LSE employs advantage of MPLS technique in order to improve scalability, QoS and ability for VPN support for Ethernet. Techniques suggested for employing MPLS function in Ethernet, such as [iii], focus only on providing L2VPN connectivity, that the methods lack functions for dynamic provisioning and differentiated service for Cho Expires - April 2004 [Page 2] LSE October 2003 each traffic flow of applications. Overhead caused from large size of frame header is another problem in particular for applications such as VoIP that generates small size of payload. However, the length of header required for supporting LSE is minimum compared to other methods because LSE uses destination address and source address of standard 802.3 header as labels. A distinct feature of LSE is that it provides service for connection establishment and resource reservation using modified ARP protocol. Pre-established tunnels are used for allocation of small flows. The edge nodes of a tunnel control admission of ARP requests and negotiate labels. Terminals use different labels (i.e. different addresses) for traffic flows of different forwarding classes. Switches that support LSE translate the labels to addresses that legacy Ethernet switches can deliver frames to appropriate LSE switches. Thus Ethernet switches do not need aware of transition of frame forwarding scheme from legacy Ethernet to Label Switched Ethernet. LSE also supports multicasting and wireless terminal mobility using a unified mechanism. Overall mechanism for LSE label assignment and frame forwarding will be explained. A soft process for evolving legacy Ethernet to LSE supporting network is proposed. 2. Classification of Network Equipments Equipments discussed in this document are classified as following. Class-1 Equipment : Legacy Ethernet equipments including NIC of user terminal. Class-2 Equipment : Ethernet switches that support LSE functions Class-3 Equipment : IP level gateways that provide services for interworking with Class-2 networks and IP networks Class-4 Equipment : Circuit based equipments that employ LSE control stacks Class-2 equipment offer services for label switched frame transmission and interworking with legacy Ethernet protocols such as spanning tree establishment. Typical class-2 network consists of class-1 equipment and class-2 equipment. Class-3 switches are IP routers providing services for both LSE label switching and IP packet forwarding. Class-4 equipments are circuit switches such as ATM or optical cross-connects (OxCs) that employ LSE protocol stack for controlling data plane. Cho Expires - April 2004 [Page 3] LSE October 2003 3. Frame Format of Label Switched Ethernet Figure-1 shows basic format of LSE frame. 0 1 2 3 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Destination Address (TID Part=24bit) | DA | +-------------------------------+---------------+---------------+ | DA (LID Part=24bit) | Source Address (TID Part = | +---------------+---------------+-------------------------------+ | 24 bit) | SA (LID Part = 24 bit) | +---------------+-----------------------------------------------+ | Length/Type (=LSE) | Payload | +-------------------------------+ | | | | | | | | | +---------------------------------------------------------------+ Figure-1 LSE Frame Format The frame format of LSE follows specification of IEEE 802.3 except that the destination address (DA) and the source address (SA) fields are used as labels. Each DA and SA field consists of 3 bytes of Tunnel ID (TID) part and 3 bytes of Local ID (LID) part. The Length/Type field indicates that the frame must be treated according to LSE protocol. 4. LSE Tunnel Establishment Class-2 switches use pre-established Ethernet tunnels in order to exchange signaling messages and establish connections. LTE (LSE Tunnel Establishment) frame, as shown in figure-2, is used for building tunnels between every class-2 switches in a class-2 network. Class-2 switches broadcast LTE frame periodically. The frame contains information of list of terminals attached to the switch, and resource demand necessary for provisioning a tunnel. Cho Expires - April 2004 [Page 4] LSE October 2003 0 1 2 3 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Ethernet Multicast/Broadcast Address (6byte) | | +-------------------------------+ | | Tunnel ID (24bit) | +---------------+---------------+-------------------------------+ | cont'd | 0x000000 | +---------------+---------------+---+----+----------------------+ | Length/Type (=LSE) |Ver|Type| reserved | +-------------------------------+---+----+----------------------+ | Root Address (6byte) | | +-------------------------------+ | | | +-------------------------------+ | | | | Policy Data | | | +---------------------------------------------------------------+ | | | Access List (Optional) | | | | | +---------------------------------------------------------------+ Figure-2 LTE (LSE Tunnel Establishment) Frame Format In LSE, TID (Tunnel ID) is used for tunnel number and LID (Local ID) is used for identification of terminals that are connected via a tunnel. The 3 bytes of TID number space is divided and allocated for each class-2 switch in a network, and so is the LID number space. A Class-2 switch that establishes a tunnel selects a TID number out of its own TID number pool. The TID number must be unique in a class-2 network. The TID number is mapped in the SA field of the LTE frame in a form of xx:xx:xx:00:00:00. Typical Ethernet switches learn SA of frames received from a port. When class-2 switches broadcast LTE frame, Ethernet switches learn the SA of LTE frames, as the frames propagating network. As a result, if a node in the network transmits a frame that DA is the SA of the LTE frame, the frame is delivered to the source class-2 switch that has broadcast the LTE frame. The data path learned from LTE broadcast is called 'Ethernet tunnel' in this document. The Ethernet tunnel is in effect a reverse path tree, rooted at a class-2 switch. Figure-3 shows an example of network that consists of 3 class-2 switches (S1, S2, S3), 2 legacy Ethernet switches (E1, E2) and user Cho Expires - April 2004 [Page 5] LSE October 2003 terminals (Ta, Tb, Tc). In the example, S3 selected a TID and composed a source address A1 for LTE frame. As a result of LTE broadcast originated from S3, network nodes learn the SA and data path to deliver frames to S3. The figure describes that Ethernet frames (that DA='A1') are delivered to S3 via the tunnel. ---------> A1 [Ta]--[S1]--[E1]---[S3]--[Tc] | /-> A1 | | [Tb]--[S2]--[E2] | --------/ Figure-3 LSE Tunnel Merging Tree Note that LTE frames contain information of resource demand (such as bandwidth, priority), and access list of IP addresses allowed for connection. Typically, the list of IP addresses are the addresses of terminals attached to a root class-2 switch, or subnet addresses that the class-2 switch provides transit service. Other class-2 switches store the information when they receive LTE frame and use it for connection control. When a class-2 switch needs to send an IP packet to a terminal (or a class-2 switch), the switch checks the stored access list and selects an appropriate tunnel. The switch encapsulates the IP packet in a frame that DA is the SA known from LTE broadcast, and transmits the frame via the tunnel. Note that the tunnel is a unidirectional path converging at a root node. Since all class-2 switches establish their own tunnels, class-2 switches are able to exchange data using tunnels. Class-2 switches use tunnels for exchanging signals and establishing connections with resource reservation. 5. Signaling & Connection Setup Most IP terminals issue ARP query when it does not know Ethernet address of the destination terminal. Using this feature, LSE provides connection oriented service in Ethernet (and also IP) without requiring substantial modification of service interface with IP layer. User terminals that support LSE request to setup a connection using modified ARP protocol. An extension of ARP packet may contain information necessary for admission control of connection request such as resource demand, authentication data, preferred policy, etc. If applications do not have ability to control modified ARP, default feature of ARP is used and connection is established without reserving resource. Such connections are used for transmitting best- effort traffic. Best-effort frames may share a connection if the destinations or default routes are identical. However, an application Cho Expires - April 2004 [Page 6] LSE October 2003 that has signaling feature, such as SIP, may directly control the modified ARP to request a connection and reserve necessary resource in a class-2 network. Figure-4 shows a timing diagram that explains procedure for connection establishment between two terminals, Ta and Tc. In the timing diagram, frame notation of 'Frame_name[DA][SA]' is used for describing type of frames and DA, SA values. Ta S1 E1 S3 Tc | | | | | | ARP[ff][Ta] | msg[G1:0][01] | msg[G1:0][01] | | | -------------> | -------------> | -------------> | assign (a1,a2) | | | | | | | | | | ARP[ff][G1:a2] | | | | | -------------> | | | | | | | | | | RPY[G1:a2][Tc] | | | | map(a1<->a2) | <------------- | | | | | | | RPY[Ta][G1:a1] | msg[Gn:0][01] | msg[Gn:0][01] | | | <------------ | <------------- | <------------- | | | | | | | | | | | | | Data[G1:a1][Ta]|Data[G1:0][0:a1]|Data[G1:0][0:a1]|Data[Tc][G1:a2] | | -------------> | -------------> | ------------> | -------------> | | | | | | |Data[Ta][G1:a1] | Data[0:a1][s1] | Data[0:a1][s1] | Data[G1:a2][Tc]| | <------------ | <------------- | <------------ | <------------- | | | | | | | | | | | Figure-4 Typical procedure for point to point connection establishment In figure-4, S1 is a class-2 switch that receives an ARP request from terminal Ta. The ARP packet contains target IP address of terminal Tc. Using the IP address, S1 examines access list to find an appropriate tunnel reaching the destination Tc. S1 encapsulates the ARP packet in a frame with additional information necessary to setup a connection. The DA (=G1:0) of the frame is the address composed using the TID (=G1) of the tunnel, and the SA is 0x01. Class-2 switches treat frames that SA is 0x01 as control messages between two edge nodes of Cho Expires - April 2004 [Page 7] LSE October 2003 a tunnel. Thus the frame is delivered to the root node S3 via the tunnel. When S3 receives the ARP packet, S3 selects two LID numbers (in the figure, a1, a2) out of its own LID number pool. The two numbers are allocated for use by terminals Ta and Tc respectively. LIDs are always assigned by root node, and the numbers must be unique in a class-2 network. Class-2 switches provide proxy service to attached terminal nodes. Terminals use proxy address assigned by class-2 switches to communicate with remote peer. Class-2 switches compose a proxy address using concatenation of TID and LID. For example in figure-4, S3 composes a proxy address [G1:a2] as a physical address of Ta and passes the ARP request to Tc. (Here, 'G1'in the brackets indicates TID part and 'a2' indicates LID part in Ethernet address) In response to this, Tc sends an ARP reply back to S3 to inform its physical address. S3 stores the information of connection (such as TID, LIDs, IP addresses, etc) in a local table, and passes the information with ARP reply to S1. The return path of the ARP reply does not necessarily identical to the path that ARP request was delivered. S1 also stores the information in a local table, and passes the ARP reply to Tc using a proxy address [G1:a1]. In summary, Ta uses [G1:a1] for proxy address of Tc, and Tc uses [G1:a2] for proxy address of Ta. Class-2 switches translate the proxy addresses to a tunnel address, and the tunnel address to a physical address of peer terminal as they relaying data frames. 6. Label Switched Frame Transmission Figure-4 shows sequences that terminal Ta and Tc transmit unicast data frames. In the middle of the diagram, terminal Ta sends a data frame to Tc using proxy address (=G1:a1). S1 splits the proxy address and use the TID part for making DA (=G1:00), and LID part for making SA (=00:a1). Since [G1:00] is the tunnel address (i.e. SA of LTE) known from LTE broadcast, the frame is delivered to S3 via the tunnel. As the frame is passing intermediate nodes (i.e. E1 and S3), the nodes learn source MAC address [00:a1] with input links. When S3 receives a frame via a tunnel, the node checks local table using LID in the SA of the frame. The local table contains information of MAC address of Tc and corresponding proxy address of peer terminal Ta (=G1:a2). S3 replaces DA to the MAC address and SA to the proxy address, and passes the data frame to Tc. In the reverse direction of frame transmission, the path learned during the upstream transmission is used for downstream forwarding. Cho Expires - April 2004 [Page 8] LSE October 2003 Note that intermediate nodes learned source MAC address [00:a1] while they pass a frame toward Tc. When the terminal Tc sends a frame using the proxy address of Ta (=G1:a2), S3 examines local table and replaces the DA to [00:a1]. S3 passes the frame to E1 because E1 is the path learned from previous frame reception. E1 also passes the frame to S1, as a result, the data frame is delivered to S1 across legacy Ethernet switches. S1 checks local table using LID in the DA of the frame and retrieves MAC address of Ta and proxy address of Tc (=G1:a1). S1 replaces DA to MAC address of Ta and SA to proxy address, and passes the frame to Ta. In conclusion, Ethernet frames are label switched using proxy addresses by class-2 switches and transmitted via legacy Ethernet switches. Traffic flows must be controlled at the ingress class-2 switches. The controlled traffic flow passing a tunnel must not be disturbed by uncontrolled traffic coming from legacy Ethernet. A priority mechanism such as VLAN can be considered in order to protect LSE traffic from non-LSE traffic at legacy switches. 7. Data Transmission in All LSE Network The method explained in chapter 6 is applied in a network where not all network nodes support LSE. A weakness of the method is that TID number and LID number must be unique in a network because legacy Ethernet switches interpret TID and LID as part of Ethernet address. In consequence, a mechanism is necessary to share available number space, and scalability is limited. When deployment of class-2 switches increase in a network and LSE switches are linked directly, it is desirable to employ a feature of Tunnel LSP (TLSP) in such part of network instead of using Ethernet tunnel. Figure-5 shows an example that all network nodes support LSE. In figure-5, a TLSP merging tree rooted at S3 is established using protocols such as LDP, or Flood Routing Protocol that is proposed in separate document. Detail of TLSP establishment is not explained in this document. --->g2 ---> g1 [Ta]--[S1]--[S4]----[S3]--[Tc] | /-> g1 | | [Tb]--[S2]--[S5] | --->g3 -/ Figure-5 TLSP Merging Tree Cho Expires - April 2004 [Page 9] LSE October 2003 The role of TLSP is similar to Ethernet tunnel. However in TLSP, TID of frames are switched as the frames pass class-2 switches. Figure-5 describes that TID (=g2) of a frame sent by S1 is switched to g1 at S4 and passed to S3. The TID numbers are set by negotiation between two neighboring nodes when TLSP is established. Since a TID number only needs to be unique in a link between two neighboring nodes, 3 bytes of TID number space would be sufficient to support most large scale Ethernet. Several distinct features of TLSP are listed below. 1) DA field of a frame is composed of 3 bytes of TID and 3 bytes of LID. The TID number in the DA is switched as the frame passes class-2 switches, however the LID number in the DA is not altered while the frame is traversing a TLSP. The SA field of a frame always indicates address of leaf node regardless of whether a root or a leaf created the frame. 2) LSE labels are bidirectional. In other words, once a node allocates TID, the number is used commonly in all neighboring nodes that share a link. This does not cause confusion even in multiple access link because LID part of DA distinguishes individual flows. LID is also used bidirectional once a root node allocates it. As a result, frames that pass a link have identical DA and SA values regardless of direction. (see figure-6) 3) When a leaf node sends a frame toward a root node (i.e. upstream), intermediate class-2 switches learn LID in the DA of a frame, and input link. The information is used for guiding frames when root node transmits frame toward the leaf node (i.e. downstream). 4) Class-2 switches are all multicast capable. Specifically, class-2 switches are able to establish multicast connections and duplicate downstream frames accordingly. Detail of multicast mechanism will be explained later in chapter 8. Cho Expires - April 2004 [Page 10] LSE October 2003 Ta S1 S4 S3 Tc | | | | | | ARP[ff][Ta] | msg[g2:01][S1] | msg[g1:01][S1] | | | -------------> | -------------> | -------------> | assign (a1,a2) | | | | | | | | | | ARP[ff][g1:a2] | | | | | -------------> | | | | | | | | | | RPY[g1:a2][Tc] | | | | map(a1<->a2) | <------------- | | | | | | | RPY[Ta][g2:a1] | msg[gn:01][S3] | msg[gn:01][S3]| | | <------------- | <------------- | <------------- | | | | | | | | | | | | | Data[g2:a1][Ta]| Data[g2:a1][S1]| Data[g1:a1][S1]| Data[Tc][g1:a2]| | -------------> | -------------> | -------------> | -------------> | | | | | | | Data[Ta][g2:a1]| Data[g2:a1][S1]| Data[g1:a1][S1]| Data[g1:a2][Tc]| | <------------- | <------------- | <------------- | <------------- | | | | | | | | | | | Figure-6 Connection establishment and data transmission via TLSP Figure-6 explains procedure for connection setup between two terminals Ta and Tc using TLSP. In LSE network, though network changes tunneling scheme from Ethernet tunnel to TLSP, terminal users need not aware of the variation. Terminal users constantly use ARP for establishing connection and proxy address for communicating with remote peer, as explained in chapter 5 and 6. In figure-6, terminal Ta initiates connection setup by sending an ARP packet to S1. The ARP packet contains target IP address of terminal Tc. S1 examines local table and finds appropriate TLSP (=g2) to reach the destination. Since S3 is the root node linked to the terminal Tc, S1 encapsulates the ARP request in a frame that DA is [g2:01]. The LID number '0x01' in the DA indicates that the frame is a control message sending to the other edge node of the TLSP. Thus the frame is delivered to S3 via the TLSP. When S3 receives the frame, S3 allocates two LID numbers a1, a2 for Ta and Tc respectively. Note that the LID numbers do not need to be globally unique because the LIDs only have local significance within a TLSP. Terminal Ta uses [g2:a1] for proxy address of Tc, and Tc uses [g1:a2] for proxy address of Ta. The rest of the connection procedure is similar to the procedure explained in chapter 5. Cho Expires - April 2004 [Page 11] LSE October 2003 The label switched data transmission using TLSP is relatively simple than using Ethernet tunnel. When Ta transmits data frame to Tc using the proxy address [g2:a1], intermediate nodes (S1, S4, S3) replace TID number of DA as they pass the frame along the path of TLSP. The intermediate nodes learn the LID number (or the whole DA) with input link, as explained in 3) above. When S3 receives the frame, the node checks local table using the LID (=a1) in the DA. S3 replaces the DA of the frame to MAC address of Tc and SA to proxy address [g1:a2], and passes the frame to Tc. When Tc replies data frame, the root node S3 replaces DA and SA to the same number as received from S4 (i.e. DA=[g1:a1], SA='S1' see figure-6 for detail). S3 passes the frame downstream to S4. Since S4 learned the LID (or the whole DA) and input link in the previous upstream data transmission, S4 passes the frame to S1 with switching the TID number in the DA. When TLSP is used, data frames received by leaf node always contain proxy address in the DA field. In figure-6, S1 retrieves MAC address of Ta using the proxy address contained in the DA of the data frame. S1 replaces the DA to the MAC address and SA to the proxy address, and passes the data frame to Ta. In LSE, although TLSP and LID number are fixed entity for relatively long period, connections are soft entity that can be erased by timer. In other words, class-2 switches must transmit a data frame toward the root of a TLSP at least once in order for intermediate nodes to learn LID and downstream path. The learned downstream path is erased by timer unless downstream terminals keep sending data and refresh it. An optimum timer management mechanism is necessary and this is left for future study. 8. Support for Multicasting and Wireless LAN Mobility The procedure explained in figure-6 is also applied in multicast connection as well as mobile connection. Note in 4) of chapter 7, class-2 switches are required to have capability to learn multicast connections. When more than two downstream nodes transmit frames of identical LID via a TLSP, class-2 switches set multicast forwarding state to duplicate downstream frames. For example in figure-5, suppose that S1 transmitted a frame that DA is [g2:m1]. S4 learn the LID (=m1) in the DA and downstream path to S1, and passes the frame to S3. When S5 transmits a frame that DA is [g3:m1], S4 merges S5 as multicast downstream path because S1 and S5 both used same LID number (=m1). When the root node S3 transmits a frame, the frame is duplicated at S4 and forwarded to both downstream nodes S1 and S5 with appropriate DA (i.e. for S1, DA=[g2:m1], for S5, DA=[g3:m1]). In summary, data frames that transmitted by a root node are duplicated and multicast along the path of TLSP, if multiple downstream nodes Cho Expires - April 2004 [Page 12] LSE October 2003 have transmitted data using identical LID number. Unicast and multicast data frames are not distinguished. An extension of ARP protocol to include semantics for multicast session is necessary. Method for selecting optimum root node (i.e. a rendezvous point) and advertising TLSP are required. Detail of mechanism for supporting multicast service will be proposed in later document. The multicasting method explained above is also used for supporting fast mobility of wireless terminals. Figure-7 below shows a situation that a mobile terminal Ta moves access point from S1 to S2 while communicating with Tc. It is assumed that a point-to-point connection has been established between Ta and Tc via a wireless link at S1. Similar to the procedure explained in chapter 7, terminal Ta uses proxy address (=g2:a1) for sending data frame to Tc. S1 puts its address in the SA of the data frame and forwards it via the TLSP. Ta (S1/S2) S4 S3 Tc | | | | | | | | | | | Data[g2:a1][Ta]| | | | | -------------> | (S1 receiving) | | | | | | | | | | Data[g2:a1][S1]| Data[g1:a1][S1]| Data[Tc][g1:a2]| | | -------------> | -------------> | -------------> | | | | | | | | | | | |request for TID | | | | | -------------> |(Ta Move to S2) | | | | (TID=g3) | | | | | <------------- | (register Ta) | | | | | | | | | Data[g3:a1][Ta]| | | | | -------------> | (S2 receiving) | | | | | | | | | | Data[g3:a1][S2]| Data[g1:a1][S2]| Data[Tc][g1:a2]| | | -------------> | -------------> | -------------> | | | | | | | | Data[g2:a1][S2]| Data[g1:a1][S2]| Data[g1:a2][Tc]| | (S1 turn off)| <------------- | <------------- | <------------- | | | Data[g3:a1][S2]| | | | (S2 relaying)| <------------- | | | | | | | | | Data[Ta][g3:a1]| | | | | <------------- | | | | | | | | | Figure-7 Data Transmission in Mobile Terminal Cho Expires - April 2004 [Page 13] LSE October 2003 When Ta moves access point to S2, Ta requests local TID number of the same TLSP tree in the course of mobile subscription process to S2. Detail of the subscription process will not be discussed in this document. S2 informs its local TLSP ID (=g3), and S1 reassembles proxy address (=g3:a1) using the new TID. When S2 receives a data frame from Ta, the node puts its own address in the SA field and forwards the frame to upstream node S4. Note that S4 still remember previous path that forwarding downstream frames to S1. Hence, S4 merges additional downstream path to S2. The data frame is passed to the root node S3, and S3 updates the location of Ta (=S2) in its local table. When Tc replies a data frame, S3 puts the address of new location (=S2) in the SA field. The data frame is forwarded via the downstream path learned by intermediate nodes. Since S4 merged two downstream paths, the frame is duplicated and forwarded to both leaf nodes S1 and S2. S1 notices that the receive frame is not designated itself because the SA (=S2) value is changed. Hence, S1 cease to forward data via its wireless link. As a result, the data frame is delivered only via the wireless link at S2. The branch path to the old access point S1 will eventually be erased by timer. 9. Structuring LSE Network When the size of class-2 network grows, a class-2 network can be partitioned according to boundary of IP subnet. In this case, independent class-2 networks are linked by class-2 switches that function as border switch. Border class-2 switches collect information of local LSP trees and advertise summarized information to neighboring class-2 networks. Other border switches that receive the information establish exterior LSP trees to relay internal traffic into the neighboring network. If class-2 networks are organized systematically, for example in tree configuration, large scale network can be built using only by class-2 switches. However, if the connections between subnets become complex, it is desirable to use class-3 switches for interconnecting class-2 networks. Class-3 switches perform label switching as well as IP routing. The switches gather information of IP routes from outside world and advertise it internally to class-2 networks. The switches also help to establish loopless concatenation of tunnels between class-2 networks using modified IP routing protocols. The switches may relay labeled frame, or terminates a tunnel and perform IP forwarding. Cho Expires - April 2004 [Page 14] LSE October 2003 On the contrary, class-4 switches are circuit based switches, such as ATM or OxC, that employ LSE control plane. Though the transmission media of class-4 switches may not be compatible with Ethernet frame, the switches can coordinate routing functions to provide long-haul connectivity between remote LSE networks. 10. Conclusion In this document, new label switching method for transmitting Ethernet frame is proposed. The mechanism of LSE is relatively simple than other proposals, and the method is efficient in terms of header overhead. There have been few proposals that incorporate services of unicast, multicast and mobility support in a unified mechanism. LSE provides solution for converging networks of different types into cost effective Ethernet. LSE multicasting has advantage of low overhead for join operation and support for bandwidth reservation. It can also support mobility of multicasting nodes. LSE improves scalability of Ethernet and provides features for establishing connections of guaranteed service quality. The method does not require comprehensive modification of upper layer interface in order to introduce connection oriented service in Ethernet. It also provides progressive evolution method to convert legacy Ethernet into LSE supporting network. 11. Security Considerations Security issues will be considered in later documents. References i Bradner, S., "The Internet Standards Process -- Revision 3", BCP 9, RFC 2026, October 1996. ii Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997 iii Kompella et. al., "Layer 2 VPNs over Tunnels", draft-kompella- ppvpn-l2vpn-03.txt, Apr 2003 Cho Expires - April 2004 [Page 15] LSE October 2003 Author's Addresses Jaihyung Cho ETRI Daejeon, Korea Phone: +82 42 860 5514 Email: jaihyung@chol.com Cho Expires - April 2004 [Page 16]