- Oct 31, 2003
3.4 Payload Mappings
So far, the multiplexing structure of SONET and SDH has been described in detail. To get useful work out of these different sized containers, a payload mapping is needed, that is, a systematic method for inserting and removing the payload from a SONET/SDH container. Although it is preferable to use standardized mappings for interoperability, a variety of proprietary mappings may exist for various purposes.
In this regard, one of the most important payloads carried over SONET/SDH is IP. Much of the bandwidth explosion that set the wheels in motion for this book came from the growth in IP services. Hence, our focus is mainly on IP in the rest of this chapter. Figure 3-5 shows different ways of mapping IP packets into SONET/SDH frames. In the following, we discuss some of these mechanisms.
Figure 3-5. Different Alternatives for Carrying IP Packets over SONET
3.4.1 IP over ATM over SONET
The “Classical IP over ATM” solution supports robust transmission of IP packets over SONET/SDH using ATM encapsulation. Under this solution, each IP packet is encapsulated into an ATM Adaptation Layer Type 5 (AAL5) frame using multiprotocol LLC/SNAP encapsulation [Perez+95]. The resulting AAL5 Protocol Data Unit (PDU) is segmented into 48-byte payloads for ATM cells. ATM cells are then mapped into a SONET/SDH frame.
One of the problems with IP-over-ATM transport is that the protocol stack may introduce a bandwidth overhead as high as 18 percent to 25 percent. This is in addition to the approximately 4 percent overhead needed for SONET. On the positive side, ATM permits sophisticated traffic engineering, flexible routing, and better partitioning of the SONET/SDH bandwidth. Despite the arguments on the pros and cons of the method, IP-over-ATM encapsulation continues to be one of the main mechanisms for transporting IP over SONET/SDH transport networks.
3.4.2 Packet over SONET/SDH
ATM encapsulation of IP packets for transport over SONET/SDH can be quite inefficient from the perspective of bandwidth utilization. Packet over SONET/SDH (or POS) addresses this problem by eliminating the ATM encapsulation, and using the Point-to-Point Protocol (PPP) defined by the IETF [Simpson94]. PPP provides a general mechanism for dealing with point-to-point links and includes a method for mapping user data, a Link Control Protocol (LCP), and assorted Network Control Protocols (NCPs). Under POS, PPP encapsulated IP packets are framed using high-Level Data Link Control (HDLC) protocol and mapped into the SONET SPE or SDH VC [Malis+99]. The main function of HDLC is to provide framing, that is, delineation of the PPP encapsulated IP packets across the synchronous transport link. Standardized mappings for IP into SONET using PPP/HDLC have been defined in IETF RFC 2615 [Malis+99] and ITU-T Recommendation G.707 [ITU-T00a].
Elimination of the ATM layer under POS results in more efficient bandwidth utilization. However, it also eliminates the flexibility of link bandwidth management offered by ATM. POS is most popular in backbone links between core IP routers running at 2.5 Gbps and 10 Gbps speeds. IP over ATM is still popular in lower-speed access networks, where bandwidth management is essential.
During the initial deployment of POS, it was noticed that the insertion of packets containing certain bit patterns could lead to the generation of the Loss of Frame (LOF) condition. The problem was attributed to the relatively short period of the SONET section (SDH regenerator section) scrambler, which is only 127 bits and synchronized to the beginning of the frame. In order to alleviate the problem, an additional scrambling operation is performed on the HDLC frames before they are placed into the SONET/SDH SPEs. This procedure is depicted in Figure 3-6.
Figure 3-6. Packet Flow for Transmission and Reception of IP over PPP over SONET/SDH
3.4.3 Generic Framing Procedure (GFP)
GFP [ITU-T01b] was initially proposed as a solution for transporting data directly over dark fibers and WDM links. But due to the huge installed base of SONET/SDH networks, GFP soon found applications in SONET/SDH networks. The basic appeal of GFP is that it provides a flexible encapsulation framework for both block-coded [Gorsche+02] and packet oriented [Bonenfant+02] data streams. It has the potential of replacing a plethora of proprietary framing procedures for carrying data over existing SONET/SDH and emerging WDM/OTN transport.
GFP supports all the basic functions of a framing procedure including frame delineation, frame/client multiplexing, and client data mapping [ITU-T01b]. GFP uses a frame delineation mechanism similar to ATM, but generalizes it for both fixed and variable size packets. As a result, under GFP, it is not necessary to search for special control characters in the client data stream as required in 8B/10B encoding,1 or for frame delineators as with HDLC framing. GFP allows flexible multiplexing whereby data emanating from multiple clients or multiple client sessions can be sent over the same link in a point-to-point or ring configuration. GFP supports transport of both packet-oriented (e.g., Ethernet, IP, etc.) and character-oriented (e.g., Fiber Channel) data. Since GFP supports the encapsulation and transport of variable-length user PDUs, it does not need complex segmentation/reassembly functions or frame padding to fill unused payload space. These careful design choices have substantially reduced the complexity of GFP hardware, making it particularly suitable for high-speed transmissions.
In the following section, we briefly discuss the GFP frame structure and basic GFP functions.
126.96.36.199 GFP FRAME STRUCTURE
A GFP frame consists of a core header and a payload area, as shown in Figure 3-7. The GFP core header is intended to support GFP-specific data link management functions. The core header also allows GFP frame delineation independent of the content of the payload. The GFP core header is 4 bytes long and consists of two fields:
Figure 3-7. Generic Framing Procedure Frame Structure
Payload Length Indicator (PLI) Field
A 2-byte field indicating the size of the GFP payload area in bytes.
Core Header Error Correction (cHEC) Field
A 2-octet field containing a cyclic redundancy check (CRC) sequence that protects the integrity of the core header.
The payload area is of variable length (0–65,535 octets) and carries client data such as client PDUs, client management information, and so on. Structurally, the payload area consists of a payload header and a payload information field, and an optional payload Frame Check Sequence (FCS) field. The FCS information is used to detect the corruption of the payload.
The variable length payload header consists of a payload type field and a type Header Error Correction (tHEC) field that protects the integrity of the payload type field. Optionally, the payload header may include an extension header. The payload type field consists of the following subfields:
Payload Type Identifier (PTI): This subfield identifies the type of frame. Two values are currently defined: user data frames and client management frames.
Payload FCS Indicator (PFI): This subfield indicates the presence or absence of the payload FCS field.
Extension Header Identifier (EXI): This subfield identifies the type of extension header in the GFP frame. Extension headers facilitate the adoption of GFP for different client-specific protocols and networks. Three kinds of extension headers are currently defined: a null extension header, a linear extension header for point-to-point networks, and a ring extension header for ring networks.
User Payload Identifier (UPI): This subfield identifies the type of payload in the GFP frame. The UPI is set according to the transported client signal type. Currently defined UPI values include Ethernet, PPP (including IP and MPLS), Fiber Channel [Benner01], FICON [Benner01], ESCON [Benner01], and Gigabit Ethernet. Mappings for 10/100 Mb/s Ethernet and digital video broadcast, among others, are under consideration.
Payload Information Field
This field contains the client data. There are two modes of client signal payload adaptation defined for GFP: frame-mapped GFP (GFP-F) applicable to most packet data types, and transparent-mapped GFP (GFP-T) applicable to 8B/10B coded signals. Frame-mapped GFP payloads consist of variable length packets. In this mode, client frame is mapped in its entirety into one GFP frame. Examples of such client signals include Gigabit Ethernet and IP/PPP. With transparent-mapped GFP, a number of client data characters, mapped into efficient block codes, are carried within a GFP frame.
188.8.131.52 GFP FUNCTIONS
The GFP frame structure was designed to support the basic functions provided by GFP, namely, frame delineation, client/frame multiplexing, header/payload scrambling, and client payload mapping. In the following, we discuss each of these functions.
The GFP transmitter and receiver operate asynchro nously. The transmitter inserts GFP frames on the physical link according to the bit/byte alignment requirements of the specific physical interface (e.g., SONET/SDH, OTN, or dark fiber). The GFP receiver is responsible for identifying the correct GFP frame boundary at the time of link initialization, and after link failures or loss of frame events. The receiver “hunts” for the start of the GFP frame using the last received four octets of data. The receiver first computes the cHEC value based on these four octets. If the computed cHEC matches the value in the (presumed) cHEC field of the received data, the receiver tentatively assumes that it has identified the frame boundary. Otherwise, it shifts forward by 1 bit and checks again. After a candidate GFP frame has been identified, the receiver waits for the next candidate GFP frame based on the PLI field value. If a certain number of consecutive GFP frames are detected, the receiver transitions into a regular operational state. In this state, the receiver examines the PLI field, validates the incoming cHEC field, and extracts the framed PDU.
GFP supports both frame and client multiplexing. Frames from multiple GFP processes, such as idle frames, client data frames, and client management frames, can be multiplexed on the same link. Client data frames get priority over management frames. Idle frames are inserted when neither data nor management frames are available for transmission.
GFP supports client-multiplexing capabilities via the GFP linear and ring extension headers. For example, linear extension headers (see Figure 3-7) contain an 8-bit channel ID (CID) field that can be used to multiplex data from up to 256 client sessions on a point-to-point link. An 8-bit spare field is available for future use. Various proposals for ring extension headers are currently being considered for sharing GFP payload across multiple clients in a ring environment.
Under GFP, both the core header and the payload area are scrambled. Core header scrambling ensures that an adequate number of 0-1 transitions occur during idle data conditions (thus allowing the receiver to stay synchronized with the transmitter). Scrambling of the GFP payload area ensures correct operation even when the payload information is coincidentally the same as the scrambling word (or its inverse) from frame-synchronous scramblers such as those used in the SONET line layer (SDH RS layer).
Client Payload Mapping
As mentioned earlier, GFP supports two types of client payload mapping: frame-mapped and transparent-mapped. Frame mapping of native client payloads into GFP is intended to facilitate packet-level handling of incoming PDUs. Examples of such client signals include IEEE 802.3 Ethernet MAC frames, PPP/IP packets, or any HDLC framed PDU. Here, the transmitter encapsulates an entire frame of the client data into a GFP frame. Frame multiplexing is supported with frame-mapped GFP. Frame-mapped GFP uses the basic frame structure of a GFP client frame, including the required payload header.
Transparent mapping is intended to facilitate the transport of 8B/10B block-coded client data streams with low transmission latency. Transparent mapping is particularly applicable to Fiber Channel, ESCON, FICON, and Gigabit Ethernet. Instead of buffering an entire client frame and then encapsulating it into a GFP frame, the individual characters of the client data stream are extracted, and a fixed number of them are mapped into periodic fixed-length GFP frames. The mapping occurs regardless of whether the client character is a data or control character, which thus preserves the client 8B/10B control codes. Frame multiplexing is not precluded with transparent GFP. The transparent GFP client frame uses the same structure as the frame-mapped GFP, including the required payload header.
3.4.4 Ethernet over SONET/SDH
As shown in Figure 3-5, there are different ways of carrying Ethernet frames over SONET/SDH, OTN, and optical fiber. Ethernet MAC frames can be encapsulated in GFP frames and carried over SONET/SDH. Also shown in the figure are the different physical layer encoding schemes, including Gigabit Ethernet physical layer, and 10Gigabit Ethernet physical (PHY) layer optimized for LAN and WAN. Gigabit Ethernet physical layer is 8B/10B coded data stream, and it can be encapsulated into GFP frames and carried over SONET/SDH. 10-Gigabit Ethernet WAN PHY is SONET/SDH encoded, and hence it can be directly mapped into STS-192/STM-16 frames.