November 2000

Network Atelier

BY KEVIN MAYER

While the next-generation is emerging, but gradually, it is also emerging everywhere, and all at once, with the deployment of each new gateway, router, switch, media server, and base station. Also no sooner do the new elements become interwoven into the network fabric than expectations rise. Can't the new network show itself in all its detail, and -- finally -- realize its grand design?

The build-out is like watching the manufacture of a tapestry. A sketch is transferred onto coarse muslin, and then weavers fill in the details. The whole workspace, the atelier, hums quietly but intensely with the weavers' industry.

Similarly, the next-generation network is becoming more finely meshed in all its dimensions, including of transport and management.

The equipment manufacturers -- the weavers -- are all eagerness to elaborate upon next-generation designs. Accordingly, many manufacturers are interested in outsourcing, avoiding the time and bother of vertically integrating their communications solutions. If outsourcing is to work, the manufacturers will have to find partners that can deliver platforms of sufficient density, and interfaces of sufficient flexibility.

Many partners are doing their best to take advantage of mass-market components and their ability to speed development, and yet create systems that merit the label "carrier grade." Such efforts are most pronounced in the area of CompactPCI. At present, we see a lot of interest in overcoming CompactPCI's limitations in terms of the telephony bus (H.110) and PCI itself. Many vendors are proposing ways to add circuit-switching capacity through segmented backplanes and bridging schemes as well as ways to relieve the H.110 bus of most of its traffic burdens, through the introduction of packet backplanes.

Higher capacity and more fully distributed configurations typically add Ethernet links to the CompactPCI backplane (currently 100Base-T, but eventually scaling up to Gigabit Ethernet). One approach includes two point-to-point links from each slot, one to each of two redundant Ethernet switch cards. Another approach, called "fully meshed," includes redundant point-to-point Ethernet links between every slot and every other slot with a full Ethernet switch on every card.

Do such innovations reduce CompactPCI's contribution to that of providing a rugged form factor? With high-speed IP or ATM infrastructures, why even have a second infrastructure like CompactPCI, which adds cost and can be a further source of system failure? One answer is that the CompactPCI bus can improve system reliability by providing a backup command and control channel and a better diagnostic mechanism for chassis management.

That it is even possible to ask such questions, and to uncover such alternatives, indicates that in some portions of the network tapestry, the threads are interweaving with considerable density.

[ Return To The Top ]
[ Return To The November 2000 Table Of Contents ]

BY NOEL LESNIAK

The key business challenges telecom OEMs face today include:

• Getting to market faster with new and innovative applications.
• Building and deploying solutions designed to run 24 hours a day, 365 days a year.
• Coping with the worldwide shortage of trained and experienced developers of HA, carrier-grade applications.

If telecom application developers are to meet these challenges, they need more, and they need it now. More to the point, they need help from their high-availability (HA) platform suppliers.

Unfortunately, many HA platform manufacturers are supplying platform solutions that are poorly supported by HA software tools. Rather than supplying HA solutions, the industry too often delivers empty HA software frameworks that require additional development work rather than fast-to-market software enablers solutions.

Aware of these difficulties, telecom OEMs are now asking their platform suppliers to help with the following:

• An off-the-shelf HA configuration and event manager software solution with rules, policies, and behaviors, plus APIs that fully support the system platform and fully support operational and network management practices used in Central Offices around the world.
• "How to" documentation, training, and support services that enable telecom HA application developers and HA driver writers to get to market quickly using the HA configuration and event management software solution.
• A certification program that assures telecom developers that the software and hardware acquired from third-party board and driver suppliers run out of the box on day one of the project and are fully integrated into the HA configuration and event manager environment.

A high-availability software architecture should include the following components:

• A Platform Abstraction Layer (PAL) that interfaces to and directly manages the underlying hardware. The PAL detects and recovers from hardware faults transparent to the operating system kernel and the applications. It signals faults to the configuration and event managers.
• An event manager that receives fault signals from HA-aware drivers, manages the removal and replacement of failed components, and communicates events to applications, the operations staff, and the network operations center (NOC). In addition, the configuration manager maintains the system's configuration and manages online changes and upgrades to it.
• Network management services that communicate to the NOC using SNMP. An SNMP agent with system MIB and event manager MIB are required. The remote center should be able to execute commands and control the local system over the network.
• IP communications services that provides heartbeat and checkpoint functions between host active and standby applications and between system functions, and can be used to transmit files between applications and to execute remote procedure calls.

APIs should be available at all levels to allow integration of applications into the HA fabric.

Telecom solution developers need a broad set of HA support services to get to market faster. Support services should include:

• A set of "how to" high-availability documentation guides for "HA aware" application developers and "HA-aware" driver writers.
• "HA-aware" applications developer and drive writer training courses.
• Development support services related to the needs of the "HA-aware" application and driver developer.

The intent of an "HA-aware" certification program is to verify that I/O slot cards and their associated drivers work properly in the HA software environment and exhibit the behaviors expected. This assures developers that their projects get off to a fast start.

Carrier-grade platform suppliers who embrace these initiatives will delight the telecom OEM by reducing their time-to-market and time-to-volume deployment. This is achieved because the HA software solutions work "out of the box" on day one, offer a complete, carrier-grade software environment on day one, and allow developers with a minimum of training and experience to be productive on day one.

Noel Lesniak is business manager of High-Availability Software Platforms, Telecommunications Business, Motorola Computer Group. 

[ Return To The Top ]
[ Return To The November 2000 Table Of Contents ]

BY HANK HENEGHAN

System designers looking to build higher performance and more reliable systems faster should be aware of a proposed new standard for CompactPCI, systems. The Compact Packet Switched Backplane (cPSB) has been submitted to the PCI Industrial Computer Manufacturers Group (PICMG) for evaluation, and its approval will mean a radically new way of designing network systems.

cPSB evolved from the confluence of new IP-based telephony applications, the growing popularity of CompactPCI and the fact that IP Ethernet switching has become the dominant LAN topology in the enterprise marketplace. It blends the robustness, reliability and hot-swap capabilities inherent in CompactPCI with the ubiquity of Ethernet -- yielding an architecture better suited for high availability and next-generation network applications. This new architecture is designed to complement existing CompactPCI systems, not replace it, thus extending the life of the rugged and familiar CompactPCI architecture.

The proposed cPSB standard is designed to move systems traffic from the shared bus architecture of CompactPCI to an embedded, fault tolerant, high speed Ethernet network. By embedding an Ethernet network onto the CompactPCI backplane, system throughput can be scaled up significantly -- from two to ten times the speed of today's CompactPCI chassis -- without affecting the legacy components or behavior of the shared CompactPCI bus. Using just 20 pins on the J3 connector per slot, the cPSB proposal leverages off the existing CompactPCI specification by overlaying two independent embedded-switched networks on the CompactPCI foundation, leaving the H.110 bus on connector J4 unaffected.

This "embedded LAN in a box" design structure can also be used to distribute communications across all 21 subsystems in one or more chassis without the use of repeaters, creating a "virtual backplane." The cPSB networks all slots together with redundant, embedded Ethernet-based IP connections, allowing cards to communicate with each other at high speed on a point-to-point basis. Each card essentially becomes a computer system to itself, off-loading a large share of the processing from the CPU card and distributing it across multiple cards -- creating a true System Area Network. If accepted as an industry standard, the proposed embedded packet-switched backplane architecture could prove indispensable to the development of the next-generation of IP products.

Hank Heneghan is senior product manager, Performance Technologies, Inc. 

[ Return To The Top ]
[ Return To The November 2000 Table Of Contents ]