Setting the Standards with Dr. Yaakov Stein

Dr. Yaakov Stein is RAD’s Chief Scientist. He writes a regular column for TechTalk in which he discusses developments in the world of telecommunications standards.

RAD invented TDM pseudowires (PWs) back in 1998, and standards have been around for several years now. Yet, I frequently receive questions regarding the different types of TDM PWs and the relationships between the different standards. In this article I will discuss the various TDM PW types (SAToP, CESoPSN, TDMoIP-AAL1, TDMoIP-AAL2), and clarify the applicable standards.

Although TDM circuits can be used to carry arbitrary bit-streams, the standards define constant-length blocks of data, which we can call structures. Familiar structures are the T1 or E1 frames of length 193 and 256 bits, respectively. By concatenation of consecutive T1 or E1 frames, we can build higher-level structures called superframes or multiframes. T3 and E3 frames are much larger than those of T1 and E1, and even larger structures are used in the GSM Abis channel.

TDM structures contain TDM data plus structure overhead; for example, the 193-bit T1 frame contains a single bit of structure overhead and 24 bytes of data, whereas the 32-byte E1 frame contains a byte of overhead and 31 data bytes. The overhead always contains the Frame Alignment Signal (FAS), an easily detectable periodic bit pattern used to delineate the structure. The structure overhead may additionally contain Operations, Administration, and Maintenance (OAM) information. We will use the term structured TDM to refer to TDM with any level of structure imposed by an FAS. Unstructured TDM signifies a bit stream on which no structure has been imposed, implying that all bits are available for user data.

Structured TDM circuits are frequently used to transport multiplexed channels. A single byte in the TDM frame (called a timeslot) is allocated to each channel. A frame of a channelized T1 carries 24 byte-sized channels, whereas an E1 frame consists of 32 channels. Note that channelized TDM must always be structured, but the converse is not necessarily true.

ATM AAL1 standards differentiated between structured and unstructured TDM methods. This terminology is somewhat deficient, because the unstructured TDM method could be used to transport structured TDM. The terminology developed for TDM PWs is more accurate.

Structure-agnostic transport refers to a non-native TDM transport mechanism that does not avail itself of the TDM structure. Such transport is ideal for truly unstructured TDM, but can equally be used for structured TDM. The important point is that any structure that might exist is ignored by the transport mechanism; in particular, no TDM framer is needed. Structure-agnostic mechanisms treat the TDM input as an arbitrary bit-stream, completely disregarding any structure that may exist in the TDM bit-stream. The structure-agnostic TDM PW protocol is called SAToP (Structure Agnostic TDM over Packet). SAToP packets contain N consecutive bytes of TDM data (N must be preconfigured and must remain constant over the lifetime of the PW), with no information about TDM structures that may or may not be present.

Structure-aware transport refers to a non-native TDM transport mechanism that avails itself of the TDM structure. Such transport inherently preserves the TDM structures, and ensures that they are properly delivered. In particular, structure-aware transport uses a TDM framer to detect the FAS, and regenerates the FAS at the other side of the emulated circuit. There are two structure-aware TDM PW protocols, CESoPSN and TDMoIP.

Structure-agnostic transport is the only option for transporting unstructured TDM, but is also suitable for transport of structured TDM when there is no need to protect structure integrity nor interpret or manipulate individual channels during transport. In particular, SAToP is the technique of choice for PSNs with negligible packet loss, and for applications that do not require discrimination between channels nor intervention in TDM signaling.

Why is packet loss significant? When a single SAToP packet is lost, an "all ones" pattern is played out to the TDM interface. This pattern is interpreted by the TDM end equipment as an AIS indication, which, according to TDM standards, immediately triggers a "severely errored second" event. Because such events are considered highly undesirable, the suitability of SAToP is limited to extremely reliable and over-provisioned PSNs.

However, when structure-aware TDM transport is used, it is possible to explicitly safeguard TDM structure during transport over the PSN, thus effectively concealing packet loss events. Structure-aware transport exploits at least some level of the TDM structure to enhance robustness to packet loss or other PSN shortcomings. Structure-aware TDM PWs are not required to transport structure overhead across the PSN; in particular, the FAS may be stripped and regenerated. However, structure overhead may be transported over the PSN when it contains OAM information.

In addition to guaranteeing maintenance of TDM synchronization, structure-aware TDM transport also can distinguish individual timeslots of channelized TDM, thus enabling sophisticated packet loss concealment at the channel level. TDM signaling also becomes visible, facilitating mechanisms that maintain or exploit this information. Finally, by taking advantage of TDM signaling, voice activity detection, or both, structure-aware TDM transport makes bandwidth conservation possible.

There are three conceptually distinct methods of ensuring TDM structure integrity: structure-locking, structure-indication, and structure-reassembly. Structure-locking requires each packet to commence at the start of a TDM structure and to contain an entire structure or integral multiples thereof. Structure-indication allows packets to contain arbitrary fragments of basic structures, but uses pointers to indicate where each structure commences. Structure-reassembly is only defined for channelized TDM; the PSN-bound IWF extracts and buffers individual channels, and the original structure is reassembled from the received constituents by the TDM-bound IWF.

All three methods of TDM structure preservation have their advantages and disadvantages. Structure-locking, used by CESoPSN, is conceptually the simplest mechanism. Each packet commences with the first byte of a TDM frame, and contains an entire frame or multiple whole frames (but this multiple must be preconfigured and remain constant for the lifetime of the PW). If there are higher-level structures (e.g. CAS superframes), these must either fit into a packet, or a special fragmentation mechanism is called into play.

Structure-indication, used by TDMoIP-AAL1, is an alternative to structure locking. Instead of requiring each packet to commence with the beginning of a frame and to contain an entire frame, a pointer mechanism is used to indicate where the next frame begins. When there are larger structures the same mechanism is used, but the pointer indicates the beginning of the larger structure (similar to ATM AAL1’s “structured with CAS” method).

Both CESoPSN and TDMoIP-AAL1 assume that the TDM channels to be transported are allocated statically and do not change during the lifetime of the PW. This behavior is inefficient for cases in which the constituent channels are often idle and need not be transported. One could save bandwidth by only placing into PW packet bytes belonging to active channels, but then an identification table would be needed, whose size could potentially exceed the savings. In ATM AAL1 systems, there was a mechanism of this sort called DBCES.

TDMoIP-AAL2 implements a more efficient mechanism, namely structure-reassembly. This mechanism functions by isolating individual channels, which are assumed to be dynamically allocated. When a channel is declared active (either by signaling or by signal processing algorithms such as voice activity detection) its contents are collected for some predetermined amount of time. To the collected data from each active channel a small header (containing the channel identifier and amount of collected data) is prepended to form a minicell, and the PW packets consist of concatenated minicells. Note that since minicells are not created for idle channels, thus automatically conserving bandwidth. In addition, since we are already collecting data from a channel for some amount of time, various signal processing mechanisms such as speech compression can be applied to further increase efficiency at the expense of processing power.

In the IETF, SAToP is described in RFC4553. The structure-aware protocols will soon become RFCs, but for the meantime are Internet Drafts called CESoPSN and TDMoIP.

Y.1413 is the ITU-T Recommendation covering structure agnostic, structure locked and structure indicated transport over MPLS. The same methods for IP transport are in Y.1453. When the network is layer-2 Ethernet, the description is in the Metro Ethernet Forum document MEF-8.

Structure reassembly for MPLS networks is described in Y.1414, and for IP networks in Y.1452.

Implementation Agreements for structure indicated, structure reassembled, and structure locked, transport over MPLS are given in MFA forum (now IP-MPLS forum) documents 4.0, 5.0 and 8.0.0.