Lean & Mean High Availability

By: Richard “Zippy” Grigonis

The microcomputer-based computer telephony era began in 1989 when Dialogic (News - Alert) introduced the first 12 channel DSP (Digital Signal Processor)-based voice processing board, the first T-1 interface board for voice processing, and the first digital TDM (Time-Division Multiplex) bus for resource sharing, the Pulse (News - Alert)-Code Modulation Expansion Bus (PEB). Boards like these from Dialogic and other vendors made it possible for developers to assemble systems for both enterprises and service providers: IVR systems, fax servers, prepaid card systems, international callback systems, inexpensive PC-based phone systems (which ultimately became the IP PBX (News - Alert)) with voicemail, auto-attendant and automatic call distribution (ACD), and call centers using PC-based predictive dialers and PC-based databases that could provide screen pops of customer data to contact center agents. However, concerns were raised over system reliability, or High Availability (HA).

It soon became apparent that readily-available, inexpensive and relatively flimsy mail order PCs were not quite up to the task of running mission critical, revenue-generating telecom applications 24 hours a day. A preponderance of plug-in CT cards would blow out a single small power supply. Round-the-clock disk drive activity would rapidly result in a dead disk drive.

Carriers, service providers and multinational enterprises were more accustomed to the “five nines” (99.999 percent) uptime afforded by the switches in the telecom network. These were based on the performance and safety requirements called NEBS (Network Equipment Building System), originally formulated by Bell Labs in the 1970s, further developed by Bellcore (now Telcordia (News - Alert)), and were made public documents in 1985. Thus, when computer intelligence began to be applied to phone calls en masse, in the enterprise, some NEBS-certified, “fault tolerant” computer systems were initially used, but such computers were quite expensive and relatively underpowered.

Around 1993, some people were becoming aware that heavy-duty “industrial computers” could be used for such computer telephony applications. These were 19-inch wide rackmounts that had passive backplanes (with 21 or so slots for plug-in boards), redundant disk storage (RAID), load-sharing power supplies and many forced air fans or “blowers”, but unlike true fault tolerant systems, they had only one CPU on a single board computer (SBC), instead of two with automatic failover capability. Unlike a true fault-tolerant computer, these devices had a single point of failure – that single CPU – which reduced their uptime to “four nines” or so instead of five. However, such a design greatly reduced the cost of the system, and thus made for reasonably-priced IVR, fax, voicemail, and other computer telephony systems.

Yours Truly, who had entered the telecom media field in 1994, pondered what to call such computers, since it was obvious they weren’t “fault tolerant”. In 1995, while in the bathroom shower one day (where all great ideas appear), I suddenly realized that Dialogic had used the term “fault resilient” to describe one of their boards. It sounded good, and so I began using the term “fault resilient computer” to describe such heretofore unclassified devices. I even wrote the book, Fault Resilient PCs. As it turned out, these devices became the workhorses that ran the applications used in the 1990s computer telephony era.

Back in 1994, when I wrote my first telecom magazine article – that just happened to be on “industrial computers” used in telecom – I covered the products of four companies, and 20 percent of their sales related to telecom. After popularizing the term “fault resilient computer” and covering the industry diligently for the next five years, I found that 80 percent of the fault resilient “bus and boards” market was now telecom-related.

A few months before I wrote my first article on the subject, in early 1994, Ziatech (later part of Intel (News - Alert), now part of Performance Technologies) and six other companies conceived of a new type of PCI-bus rackmount that could be suited for the telecom market. It was electrically similar to the PCI bus in desktop computers, but was ruggedized for telco-related use and supported component hot swappability. Ziatech called it “RuggedPCI” and the specification was limited to the 3U-high form factor. At a September 1994 meeting of PICMG (News - Alert) (PCI Industrial Computer

Manufacturers’ Group), Dennis Aldridge, a marketing director at Texas Microsystems (now part of RadiSys (News - Alert)) didn’t like the name “RuggedPCI” since his passive backplane products were already “rugged”. Joe Pavlat, then running a company called Pro-Log (later part of Motorola), on-the-spot came up with the “temporary” name CompactPCI (News - Alert), which actually became permanent. Wayne Fischer of Force Computers (now part of Motorola) then began using the abbreviation “cPCI” in his emails and other correspondence. It too became popular.

CompactPCI has had a jolly good run, but telco demands for an even more powerful, high density, and super high availability computing platform led PICMG to develop today’s Advanced Telecom Computing Architecture (better known as AdvancedTCA (News - Alert), or simply ATCA) form factor machines.

One of the first companies to take ATCA seriously was RadiSys, which continues to offer its potent Promentum systems. For example, the hefty, carrier-grade, 13U-high Promentum SYS (News - Alert)-6016 supports 10 gigabit-per-second (Gbps) switching, packet and media processing over its 16-slot backplane. It comes standard with the Promentum ATCA-2210 10-Gigabit switch and control modules that provide integrated centralized functions such as switching, shelf management, network-timing and system management capabilities. The 6016 is used in 3G wireless and wireline infrastructures for apps such as Radio Network Controllers (RNC/BSC), Media Gateways, IMS (CSCF, Application and Media servers), and IPTV (News - Alert).

The SYS-6016’s NEBS Level 3-ready, highly redundant architecture eliminates any single point of failure and enables 5 nines or even 6 nines availability.

To power AdvancedTCA systems at maximum capacity, take a look at RadiSys own ATCA-4500 10-Gigabit Compute Processing Module. A 7th generation single board computer from Radisys, the ATCA-4500 is a single slot AdvancedTCA computer module based on the awesome, single socket L5518 Intel Xeon processor.

The “Middle” In HA Middleware

RadiSys has always been at the forefront of developing systems of high availability. Back in 2006 they partnered with GoAhead software on a standards-based, pre-integrated high availability ATCA solution, an integration of a RadiSys Promentum with GoAhead SelfReliant, standards-based high availability middleware. Later in that same year they partnered with OpenClovis (News - Alert) to resell and support OpenClovis high availability middleware and application development tools with the Promentums. By pre-integrating ATCA platforms with such high availability middleware, equipment providers could now focus on building interesting differentiated applications rather than squandering resources developing basic protocols and platform management solutions themselves.

GoAhead Software (News - Alert), of course, is one of the founders of the Service Availability Forum, a consortium of communications and computing companies working together to develop and publish HA and management software interface specifications.

In the ‘Open Source’ column elsewhere in this issue, Yours Truly interviews Dr. Asif Naseem, President of the SA Forum and President and COO of GoAhead Software, which recently released its SAFfire 3.0 high availability and management middleware capable of supporting the open specifications of the SA Forum. The platform-independent SAFfire supports the SAForum’s Application Interface Specification (AIS) and the Hardware Platform Interface (HPI (News - Alert)).

With SAFfire 3.0, GoAhead has mastered integration with such hardware platforms as the Continuous Computing FlexTCA platforms, the IBM (News - Alert) BladeCenter family, and RadiSys Promentum ATCA chassis and blades. SAFfire 3.0 is also pre-integrated with leading software solutions such as Tail-f Systems ConfD network configuration management, Wind River Linux and VxWorks and MontaVista (News - Alert) Linux operating systems, and the Oracle TimesTen in-memory database. Also, SAFfire interoperates with the IBM solidDB in-memory database.

Moreover, the new GoAhead Service Availability Forum Ecosystem (SAFE) Program is a partner program aimed at reducing the time required for equipment manufacturers to deliver mission critical systems and applications. By teaming with major providers of open standards-based, Commercial-Off-The-Shelf (COTS) hardware and software, GoAhead strives to eliminate the risk and cost associated with platform integration and validation.

GoAhead recently named John Hansen as their new CEO and Chairman.

As for OpenClovis, their work on high availability middleware continues to ride the 10+ year wave of open source and COTS technologies infiltrating the world of Telecom Equipment Manufacturers (TEMs) and Network Equipment Providers (NEPs). (Carrier-grade Linux, for example, as well as OpenClovis’ own open source HA middleware.) Their OpenClovis Application Service Platform enables developers to quickly add a management and high availability service layer to any telecom design, thus enabling highly marketable, differentiated solutions.

Furthermore, their OpenClovis IDE (Integrated Development Environment) software solution, when used in conjunction with the OpenClovis platform, streamlines the specification of a system’s information model, HA aspects and the communication infrastructure. OpenClovis IDE stores all information describing a project in XML files modifiable by the user. Modeling of system resources and relationships are specified using a graphical UML editor. IT

Richard Grigonis (News - Alert) is Executive Editor of TMC’s IP Communications Group.

GoAhead Software - (www.goahead.com)

OpenClovis - (www.openclovis.com)

RadiSys - (www.radisys.com)

Service Availability Forum - (www.saforum.org)

» Internet Telephony Magazine Table of Contents