Keeping it open

Jan 9, 2002 12:00 PM

Despite the recent abrupt slowdown in telecom infrastructure investment, focus on elements for the converged network, also known as the next-generation network (NGN), remains strong. This is driven primarily by the potential for new services that place voice and data capabilities on the same connection simultaneously, but also by the lure of lower operating and maintenance costs for the carriers. Furthermore, the continued growth in wireless (and the march towards 3G wireless networks in particular) adds more spice. So, at a time when the downturn in the telecommunications market has forced major equipment vendors to shed employees and reduce costs, the need to develop new competitive solutions remains as strong as ever.

To develop these new competitive solutions, many telecom equipment vendors are accelerating the evolution of a new business model by choosing to outsource development and use off-the-shelf open standards equipment rather than develop everything in house. However, to a large extent, the open standards industry has been most successful in small- and medium-density installations where time to market and high levels of functionality are key. The industry historically has been less successful in reaching the reliability and density required by a big carrier switch such as the Nortel DMS100 or the Ericsson AXE. This article examines some of the demands on the "big switch" elements in NGNs and how the open standards industry is evolving to meet the challenge.

How big does it all need to get?

The sizing and network topology for switched circuit network elements differs between service operators, but a local (Class 5) exchange could be from 2000 to 20,000 to 100,000 ports while a tandem (Class 4) exchange in a large network could be many hundreds of thousand ports. The demand is steadily growing, but the space to host exchanges is not.

Service operators see floor space and power as one of their major overheads and are unlikely to accept much increase in floor area to move to NGN. In fact, a reduction over the equivalent switched circuit network version would be better received. The new generation of circuit switches is increasing in density to compensate, so the bar for the NGN continues to get higher. In fact, there is a continuous demand to increase density while reducing power consumption. Basically, NGN must offer demonstrable operational savings and fast time to profit in order to pass the Finance Director!

This means that, ideally, trunking gateways must achieve port densities of between 40,000 and 80,000 ports per frame to compete with carrier switches -- a frame varies according to company but can be considered roughly 23 inches (600mm) wide and 7 feet (2.1m) or 8 feet (2.4m) high. The minimum processing requirement for each active channel is voice packetization and echo cancellation (up to 128ms echo tail). Access/edge gateways must do this and also potentially generate dial tone, play prompts, collect digits and sometimes perform additional voice compression between the backbone network and the low-bandwidth user connection as well. A gateway onto the UMTS (3G cellular) or the GSM network must provide voice compression to the appropriate wireless standard. For example, GSM-AMR, the new adaptive multi-rate codec used in GSM, can compress from standard telephone speech at 64 kb/s down to 4.75 kb/s.  Where compression is used, the system must also find ways to transport modem, fax and even DTMF digits in an alternative path because they cannot pass through the compression stage and survive.

Can open standards implementations pass muster?

A lot has been written about CompactPCI equipment practice as the standard for telecom equipment. CompactPCI is a standard owned, maintained and improved by the PCI Industrial Computer Manufacturers Group (PICMG). For telecom, CompactPCI is typically based on a Eurocard 6U (233mm high by 160mm deep) plug-in board. When mounted vertically, the inter-board spacing means it is possible to get 21 boards into an industry standard 19-inch wide shelf. Adding power conversion, cooling fans, etc., means that up to four typical high-reliability CompactPCI shelves can fit in a frame.

Therefore, to achieve the target of 80,000 ports per frame, each shelf must reliably handle 20,000 ports -- an average of 1000 ports per board. That needs to include packet network and circuit network connections as well as the processing needed for the voice packetization. Throw in the need for "five-nines" reliability as a minimum (99.999% availability, which represents five minutes of planned and unplanned downtime in any year) with a strong desire to move to "six nines" (99.9999% representing 30 seconds of downtime), and the scale of the challenge starts to become clear.

Certainly multiple DS-1 (T-1/E-1; 24/30 channel) class voice packetization boards have no play -- even DS3 (T3; 672 channel) boards are pushed hard. The real need is for optical STM-1 (2016 channel) or STM-4 (8064 channel) class boards, and the good news is that the core devices needed to make this possible are now becoming available. That way, with N+1 redundancy, additional boards for call control, element management, alarms, etc., 20,000 ports per CompactPCI shelf starts to become a realistic possibility. The issues to solve are more to do with reliable and scalable data transport within the shelf and reliable data aggregation and distribution at the edges of the shelf with appropriate protective switching.

CompactPCI has been evolving to take account of these trends. The recent PICMG 2.16 standard moved a dual redundant star gigabit Ethernet network into the backplane, thus removing a lot of cabling problems. Full gigabit switches will follow shortly. The PICMG 2.17 standard (still in work) offers the means to extend the current timeslot bus well beyond the 2048 port limit imposed by the existing PICMG 2.5 (H.110) standard. By far the biggest recent advance, however, is in the data arena. High-speed ATM and storage data capabilities have been extended with the announcement of the PICMG 2.20 CompactPCI Serial Mesh Backplane (CSMB) standard proposal. CSMB is capable of more than 700 Gb/s data switching capacity. This uses a point-to-point connection matrix in the backplane (the mesh) so that each board has a 1:1 bi-directional link with every other board.

Still, CompactPCI has its limits. The growing demands on the board/backplane interface have reduced the effective area and power budget for doing useful work. Accordingly, PICMG is starting to consider a completely new set of standards for plug-in blade based equipment called PICMG 3.x. These new standards are being developed to meet the needs of the core network elements the industry is considering. The boards will be bigger, have better cooling to support much higher power consumption, and feature a full mesh data backplane similar to CSMB. The first-draft standards based on this initiative are aggressively scheduled for publication in Q3 2002, and with more than 100 companies participating in the recent kick-off meeting, it promises to be an interesting time.

What if we don't want to change all the equipment practice?

For some companies looking to evolve existing equipment, the use of CompactPCI or other standards represents a total break with previous equipment practice and is impossible to justify. Fortunately, these companies have another option -- the use of a standards-based mezzanine card on top of their own specially designed carrier card. The most popular mezzanine standard for this is PMC (PCI Mezzanine Card). Mechanical and electrical attributes are defined in IEEE1386.1, and extensions defining pin assignments for telecom applications have been recently approved by PICMG as PICMG 2.15 (also known as PTMC - PCI Telecom Mezzanine Card).

It is possible to fit multiple PMC modules onto carrier boards that then plug into the proprietary backplane and chassis used in previous equipment generations. The PMC modules can provide additional processing, line or network interfaces, and high-density voice packet processing functions. Again, prime driving issues are power and channel density but with continual advances in DSPs and network processors, this is certainly a viable option today.

Scalability -- the watchword for rollout

We have discussed the scalability of the actual core network equipment in terms of the media gateways needed to carry voice traffic. The call control needs to be similarly scalable -- handling millions of call attempts. The classical view of the NGN has all the call control vested in an element called a softswitch with the media gateway acting as an event collector and relay. The media gateway only performs bearer signaling where it is required to be carried in the same physical interface as the bearer channels (such as CAS). In other cases, the media gateway passes all call state information to the softswitch.

Softswitches are principally large transaction processing computers that need to meet the same scalability and reliability demands as the other elements. Advances in server architectures based on loosely coupled processor blades running high-availability and multi-processing extensions are finding a home here. Because all call flow runs through the softswitch, service operators can quickly generate and change call flow applications, even open the interface to additional application developers.

As open as it is intended to be, this centralized control model goes against the most successful data network paradigm -- that of the Internet. The Internet has evolved to the extent it is today by having a relatively simple routing protocol driven by intelligent clients. As such, some believe the NGN call control model is flawed and that more of the call control and signaling intelligence must be pushed down to the access points and even to the client devices with distributed protocols like SIP coming to the fore. Consequently, we begin to see intelligent client devices such as IP telephones begin service with early adopters, and the 3GPP have adopted SIP as their client signalling protocol for future 3G handsets.

However, there are billions of standard DTMF telephones and mobile handsets in the world, and they will not be replaced overnight. The first point at which that these meet the NGN is likely to be in a media gateway of some sort. Perhaps the extra scalability lies in making the media gateway do more of the call signaling processing for both packet network and switched circuit network without step-by-step reference to a softswitch. It remains to be seen whether one or the other, or a mixture of both, wins in the end.

Brian Carr is a telecom product manager for the Motorola Computer Group.  He can be reached at briancarr@motorola.com.

Visit Motorola online.

FYI...

OPEN, REDEFINED
Nov 19, 2001, Telephony, by Vince Vittore
Nokia's announcement last week that it will lead the creation of open standards for mobile devices to encourage...

Pace adds features, functions to home gateway product
Nov 12, 2001, TelephonyOnline.com, by Jim Barthold
Pace Micro Technology Americas has introduced an IPTV digital home gateway set-top box that upgrades its existing DSL streaming media delivery system...

Nortel, IBM to collaborate on enterprise IP
Oct 25, 2001, TelephonyOnline.com, by Jim Barthold
Nortel Networks and IBM Global Networks are expanding their existing relationship to include Nortel’s open-standards-based integrated voice-over-IP [VoIP]...