April 1999

Open Source For CompactPCI

BY BROUGH TURNER

CompactPCI is a success story that is still unfolding, but it is unfolding quickly, enlarging the theme of open telecommunications. Just over three years ago, CompactPCI was first specified by the PCI Industrial Computer Manufacturers Group (PICMG). Even more recently, just one year ago, PICMG defined a telecom-focused variant of CompactPCI.

While these are recent developments, CompactPCI has already attracted a following, displacing the venerable VME standard in most new industrial computer "design wins." In addition, it has penetrated each of the major telecommunications equipment manufacturers - companies like Lucent, Nortel, Siemens, Ericsson, Motorola, and Nokia - in each case, displacing proprietary designs for one or more new projects.

The emergence of CompactPCI represents a turning point in the design of telecommunications infrastructure equipment. Because CompactPCI runs the same software as desktop PC systems, telecommunications equipment developers can now leverage the world's richest software development environment, including the wealth of components available for mass-market computing systems. At the same time, CompactPCI's robust architecture, with its support for the "hot swap" of boards in a live system, helps CompactPCI address the reliability and serviceability requirements of the telecommunications market.

OBSTACLES TO OVERCOME
Based on design wins, the CompactPCI market will be much larger than the VME market - but when? The joke today is that CompactPCI is a zero billion dollar market! Everyone knows it will be a multi-billion dollar market eventually, but today sales are considerably less than $100 million.

During 1998, many CompactPCI telecom products were announced, and some began shipping. But, as Natural MicroSystems found out when attempting a beta test of their first products in June 1998, it was difficult at that time for customers to obtain CompactPCI chassis that met both the telecom and hot-swap specifications. Later in 1998, that difficulty had been resolved. Multiple vendors had telecom chassis available. But, by the end of 1998, a new challenge appeared.

The new challenge again pertained to hot swap, but this time the issue was software, not hardware. On the hardware side, PICMG defined the CompactPCI Computer Telephony Specification, incorporating the H.110 telecom bus specified by the ECTF. This specification defines the hardware, and the functional sequences for hot swap and operation of the telecom bus, but there is no comparable specification for the low-level software needed to make hot swap, or indeed telecom circuit switching, work.

There are many, perhaps too many, standards for high-level telecom functionality (TAPI, JTAPI, S.100, etc.), but there are no standards for the lower level software needed to take advantage of the new CompactPCI telecom hardware. Many vendors have announced boards that are CompactPCI and H.100 bus compatible, but the software drivers and lower level services necessary to make either hot swap or telecom circuit switching actually work are lagging behind by as much as a year or more.

Some software work for CompactPCI hot swap has been funded by PICMG, and there are development efforts underway by operating system vendors such as Microsoft, Sun Microsystems, and SCO. But the PICMG-funded effort is for Windows 2000, an operating system still in beta test. And the work going on within the operating system vendors is focused on future releases of their operating systems. There are individual capabilities, available from a few specific companies, but standard software support for hot swap is lacking, and this deficit is delaying system developers, postponing the day when we realize the promise of CompactPCI.

For example, I am aware of a major telecom equipment manufacturer with a project underway for a new generation of wireless telecom equipment. For this project, Motorola Computer Group won the CompactPCI system business, Natural MicroSystems won the business for DSP and telecom line interface boards, and Brooktrout's Netaccess division won the frame relay interface business. But the customer needed working hot-swap software. To keep this project on track, Natural MicroSystems ended up offering their CompactPCI hot-swap software to Brooktrout.

Collaboration of this sort is hardly unique. Many of the more attractive opportunities require a level of cooperation between companies that may otherwise be competitors. And, by late 1998, Natural MicroSystems was contemplating several software licensing deals.

AN IDEA IS FORMED
As a result of several experiences like the one above, Natural MicroSystems decided that it would make sense to just give away their CompactPCI hot-swap and circuit-switching software. They felt it would be to their commercial advantage to accelerate the growth of the CompactPCI market in the telecommunications industry - even if it meant giving away software that they had developed at considerable expense.

Early in 1999, Natural MicroSystems held conversations with several CompactPCI component suppliers and telecom system developers to determine if they would adopt the Natural MicroSystems software if it was made available on a more public basis. The response was overwhelmingly positive. In addition, there were many specific suggestions about what to include and how to release it. But the preferred approach, it was clear, was to follow an open source development model. (See this column's March 1999 installment, entitled "Open Source Software Infuses CTI".)

TURBO-CHARGING THE GROWTH OF CompactPCI
After reviewing other open source projects, Natural MicroSystems decided to follow the open source model developed by Netscape for the public release of their browser code. This is the same approach taken by Ericsson Telecom in the release of their Erlang distributed middleware for highly available systems. Upon hearing of the Natural MicroSystems plan, other companies offered to endorse the effort, several companies offered to contribute additional software, and the PICMG board of directors offered to host the web site.

The Open Source for Open Telecom initiative rolled out in early March with a web site at www.opentelecom.org. This web site includes Natural MicroSystems' source code for their Hot Swap service; their Point-to-Point Switching service; the low-level command and event dispatcher that forms the core of their CT Access telecom software; and the device drivers for several Natural MicroSystems boards - all under a variety of operating systems (Windows NT, Solaris, UnixWare, and others).

The device drivers are board-specific, but this source code should help other companies that want to port hot-swap and switching software to their board-level products. In addition, Lucent Technology and Telgen, Inc. have contributed compatible source code for the switching device drivers for the demonstration boards that Lucent Microelectronics sells for their T8100 series of H.110 switch chips. The Telgen software includes demo code with a convenient user interface under Windows NT.

The software on the opentelecom.org web site is field-tested code. The hot-swap software is code that was first demonstrated over a year ago and has been widely available for six to eight months. The circuit-switching software for H.100 and H.110 has been shipping even longer. In addition, the circuit switching software benefits from work originally done in the MVIP community to support diverse vendors' boards. The current software fully supports the ECTF H.100/H.110 bus, which is an interoperable superset of the MVIP-90 bus, H-MVIP bus, and Dialogic's SCbus. The circuit-switching software also includes a standard driver software model that is known to work with a wide variety of circuit-switching bus interface chips and specific software implementations for chips from Lucent Microelectronics and Mitel Semiconductor.

While device drivers are typically operating system-dependent, Natural MicroSystems has abstracted the operating system dependencies out of the rest of the Hot Swap and Point-to-Point Switching services into a single module. The opentelecom.org web site contains sample versions of that module for several operating systems, including Linux, and this organization should aid those desiring to port to real-time and embedded operating systems.

At least three companies are offering commercial contract support services that may be useful to those wishing to utilize this code base - TelGen, Pigeon Point Systems, and Natural MicroSystems. TelGen has experience with both hardware and device drivers for CompactPCI and the H.110 telecom bus. Pigeon Point Systems has done hot-swap software development on contract for Natural MicroSystems, PICMG, and others. And Natural MicroSystems' Services business unit is able to provide custom support services for projects using this code base.

BENEFITS ABOUND
Support for the Open Source for Open Telecom initiative has been overwhelming, as there are a wide variety of companies who benefit. Component suppliers, board vendors, and the developers who use their components and boards, gain a time-to-market advantage because they do not have to create this software from scratch. This gives them more time to focus on their core values.

System developers gain a similar time-to-market advantage because they have ready access to the low-level "plumbing" that integrates boards from multiple vendors and allows them to provide point-to-point circuit switching between telephony end points on different boards in a common chassis. Service providers also benefit because, as the CompactPCI industry grows, they will be able to deploy new revenue-generating products and services much more rapidly.

MOVING TO THE NEXT LEVEL
The computer telephony industry, and its use of open systems, has enjoyed great success in the corporate enterprise. Opening up the rest of the telecommunications industry, including the public telecom network infrastructure markets, has taken longer. But, it is well under way. Open systems have been accepted by service providers as a means to rapidly develop enhanced service applications - necessary if they are to differentiate themselves from their competition. And CompactPCI, with its high availability and serviceability features, is key to making open solutions that meet the stringent service availability criteria of the telecommunications industry.

The Open Source for Open Telecom initiative will help spread the key software components that are needed to reap the benefits of CompactPCI for telecommunications. By combining a proven code base with an open source development model, this software should continue to evolve, and to make CompactPCI the standard that fuels the growth of high-capacity, high-performance, network-based open telecommunications platforms.

Brough Turner is senior vice president of technology at Natural MicroSystems, a leading provider of hardware and software technologies for developers of high-value telecommunications solutions. For more information, call Natural MicroSystems at 508-620-9300 or visit the company's Web site at www.nmss.com. E-mail to the author ([email protected]) is also welcome.