Next-generation SDH — it's just IP over WDM


(Note: In this article the term SDH is used in equivalence to SONET).

'Next-Generation SDH" is a term embracing several new ITU-T SDH standards that have emerged, firstly, to make SDH better fitted to transport the increasing amounts of packet data in the network and, secondly, to implement a signalled control plane that enables fast and automatic provisioning of transport services.

A fruitful cross-pollenation has taken place between the IP and transport worlds, where problems highlighted by the rapid growth of data have forced new solutions. At first these were seen as competitive but they have since been worked through and have now emerged as SDH standards fitting both circuit- and packet-oriented systems.

Although SDH was designed to carry data, it was developed in a pre-Internet world. When the packet traffic started to grow at 100% per year, compared with a few percent per year for voice traffic, some strains began to show in the system. SDH's many shortcomings were therefore a topic of much industry discussion.

Re-allocation of bandwidth was achieved through re-provisioning by a centralised management system, often including a complete, manual request and purchasing process. This did not fit well with the dynamic and unpredictable nature of high-capacity data flows.

When data was carried on SDH, it was usually mapped by protocols designed for other applications, or pre-standard, such as PPP and POS. New protocols became important — Ethernet, FICON, and ESCON.— and a plethora of different and proprietary mappings posed compatibility problems.

In the original SDH multiplexing structure (2/8/34/140Mbit/s), the four-fold increments became too large for optimal provisioning of packet data flows. The maximum service rate was too low; the highest-rate virtual container (VC-4, 140Mbit/s) became too small for the higher bandwidths used in data flows, Ethernet and router ports. The virtual concatenation of VCs, though standardised, was not widely used.

There was no direct routing of 2.5Gbit/s or 10Gbit/s (express layers). As bandwidths grew, it became uncomfortably expensive to demultiplex all lower-order streams that did not need to be re-routed in cross-connects.

There was no interaction with higher layers. All layers, not only SDH, operated in isolation. SDH, ATM and IP were all separate networks, and only by human interaction were resources such as bandwidth or protection moved or utilised between layers.

These issues stem from the higher bandwidths and unpredictability of the data flows, requiring both higher data rates as well as much higher and automatic flexibility in the network. It was clear that no single layer would fix the problems.

Many new ideas were brought to the ITU to standardise the various improvements on SDH that were proposed. Eventually, in 2001, a set of new ITU-T standards emerged in their first versions. Of these, the most practically important are at present:

For a long time SDH has had the ability to divide a payload over several virtual containers (concatenation) described in G.707. This can be done with VCs either forced to be adjacent in time (contiguous concatenation, VC4-4c) or associated but sent over various paths and then re-ordered at the receiver (virtual concatenation, VC4-4v). Contiguous concatenation preserves timing, but it may not be possible to find the required number of adjacent, empty VCs in a given link, whereas virtual concatenation can assemble isolated free VCs. The timing is uncertain since they may traverse different paths. This feature has not been so much used, but it fits very well with less timing-critical data applications.

To this has now been added:

This is quite a short document which builds on the existing virtual concatenation. It defines the automatic set-up and tear-down of virtually concatenated VCs so that a variable-bandwidth pipe can be generated which can be adapted in steps (2Mbit/s to 140Mbit/s) to fit the payload bandwidth. The signalling is done by the normal SDH Network Element and Network Management systems, using the SDH overhead.

An elegant feature of this scheme is that the payload is automatically mapped into the available VCs even as they change in number, up or down. This means that the bandwidth adjustments are hitless. It also means that the LCAS scheme has a soft protection mode, where the link does not go down completely in case of failure. If only a few of the VCs are lost due to failure of some link, LCAS automatically re-adjust the payload between them.

Once the variable bandwidth pipe has been created, a standardised way is needed of mapping data flows of varying origin and format into the VCs of the pipe. A Generic Framing Procedure (GFP) is therefore defined in Recommendation G.7071.

This procedure exists in two flavours: one frame-mapped (GFP-F) and one transparent (GFP-T). The frame-mapped version is aimed at transferring complete client frames, such as for Ethernet/MAC or IP/PPP. This means that the frame must be assembled in its entirety before transmission, which may cause propagation delays. It therefore requires, for example, buffering and MAC awareness, and is now specified only for Ethernet.

The Transparent mode is aimed at direct transmission of client data streams with requirements of low latency such as Fibre Channel, ESCON, FICON, or Gigabit Ethernet. This mode operates on any 8B/10B block-coded client data stream, and sends it over fixed-size GFP frames using a 64B/65B block code.

Thus, one generically applicable mapping standard can now replace the multitude of ways to map data into SDH frames. Moreover, GFP is not restricted to mappings into "classic" SDH VCs, but can also be applied to the higher-capacity (2.5Gbit/s to 40Gbit/s) pipes obtained from Recommendation G.709 (below).

The largest of the new Recommendations is G.709, which defines several new interfaces, frames and functions. In particular, it introduces higher-data-rate-containers up to 40Gbit/s, forward error correction (FEC), and it operates on multiple wavelengths. In short, it is a sort of "super-SDH frame".

The core of G.709 is the new frame structure capable of containing any client, such as Ethernet, IP, or complete unmodified STM-n-frames (that is, SDH as a transparent "client" to itself) augmented by advanced FEC taken from earlier submarine standards, and a combined optical/electrical multiplexing hierarchy.

Each client signal is first mapped into an Optical channel Payload Unit (OPU) that serves to adapt the client to the main transport frame, the Optical channel Data Unit (ODU; see Fig.2). This is, in a sense, the new equivalent to the VC-4. The OPUs and ODUs also have an order (e.g. ODUk), with k=1 meaning 2.5Gbit/s, k=2 10Gbit/s, and k=3 40Gbit/s line-rates.

Up to six ODUk can then be connected in tandem, before FEC is applied to the aggregate, and overhead added, creating an OTUk.

To the OTUk is then added optical channel specific overhead, and this constitutes the OCh, i.e. the Optical Channel itself.

Several (n) optical channels (wavelengths) can then be aggregated into an OMU-n.m (Optical Multiplex Unit) which, together with multi-channel overhead information, constitutes the final OTM-n,m (Optical Transport Module). For the optical channels, the order (k) is denoted by m.

For example, the signal from a 32-wavelength 10Gbit/s system including an optical supervisory channel, and conforming to G.709 would be described as an OTM-32.2 signal. Compare STM-64 for a single-channel 10Gbit/s SDH signal.

These ODUs are already being implemented as switching levels in emerging super-band cross-connects, which will thus be able to directly switch high-order data flows without decomposing them into low-order VCs. Moreover, the OCh, OMU, and OTM overheads contain management for several wavelengths, thus also providing functions for future all-optical switching. Finally, the ODUs can be used together with GFP to directly map data, even in 10 or 40Gbit/s streams into the new frame structure, thus making the Next-Gen SDH standards an enabler for either "IP over WDM"-style networks or "carrier-class Ethernet".

There is much more on-going co-operation between data and telecom network standards, to be described in another article. The ITU is also working on several standards focusing on the automatic provisioning part of future networks, known as ASON — the Automatically Switched Optical Network.

Together with the GMPLS (Generic Multi-Protocol Label Switching), via the IETF (Internet Engineering Task Force), and open optical interfaces such as the UNI (User Network Interface) and NNI (Node Network Interface) from OIF (Optical Internetworking Forum), an automatic signalling system is envisioned for the transport network.

This will replace today's manual provisioning and management with autonomous signalling, thus reducing network operation cost and speeding up service provisioning to the point that, for example, a client IP network can directly interact with the transport network when it needs a change in service (bandwidth, protection, QoS etc).

But that is another story. Suffice it to say that the rumour of the demise of SDH is greatly exaggerated — in fact, it is rapidly becoming just the missing link in the development of IP over WDM systems.

Per Andersson
Director, Business Development
Dr Per O Andersson is director of Business Development at Ericsson (Transmission &.Transport Networks).

More in Packet Transport