When Intel, Motorola, Siemens, NEC, Nokia, Lucent, Alcatel and Nortel get excited about working together on a technology, everyone—from software manufacturers to venture capitalists—takes notice. That is the case with the Advanced Telecom Computing Architecture (AdvancedTCA, www.advancedtca.org), a series of specs that are drawing together the once disparate worlds of telecom and computing.
Because the “buy-don’t-build” craze of the ‘90s led telecom management to stick non-standardized systems into their computing environments, thousands of pages of engineering documents were created, tweaked, convoluted and in many cases, lost. With the telecom meltdown, the engineering groups that created those voluminous documents dissolved or shrunk, leading to major operational and maintenance headaches.
In 2001, a group of supply chain management employees from Lucent approached the PCI Industrial Computing Management Group (PICMG), a consortium focused on standards for ruggedized equipment in the computer industry, and unintentionally planted the seed for what is being called the biggest evolving technological revolution to hit telecom since the PC hit the computing industry.
“Initially, we never thought of using any of our specifications for central-office grade applications,” says Joe Pavlat, president and chairman of PICMG. Since then, PICMG 2.0 has evolved into the Advanced TCA 3.0 that exists today.
While ATCA specs may not sound sexy in their most basic definition, their impact is predicted to be huge for carriers trying to avoid forklift upgrades around proprietary equipment.
ATCA has re-energized networking companies and computing companies, which are quickly becoming aware that carriers and ISPs want the freedom to plug-and-play, rather than remain beholden to one manufacturer. “There will be a 3.5 percent increase in spending for new equipment over the next four years, which means carriers will spend approximately $200 billion on ATCA-compliant equipment—expected to represent at least 10 percent of total equipment purchases by 2006,” says John G. Metz, principal analyst, Metz International, which is focused primarily on ATCA-based activities. That fact has reenergized the OEM and computing markets, as blade server start-ups with flexible components could compete with the biggest OEMs should ATCA gain traction. Potentially, high operating expenditures around stocking large arrays of servers, blades, cards and other hardware could become a thing of the past. “CIOs will stock one-tenth of what they have to stock now, thus reducing OpEx by 25 percent by 2006.”
Standardization would mean IT organizations wouldn’t have to staff engineers solely for their knowledge of proprietary components, as ATCA-compliant DSLAMs, terabit routers, and everything in-between, will be downloaded onto blade servers.
Billing and OSS vendors could run software on blade servers from any manufacturer, utilizing myriads of CPUs and MIPS processors—both of which can be used as blade servers.
While OSS and billing software can still run on traditional servers, they will integrate easily with blade servers.
“If you want to add DSL services, VoIP or IP traffic, you just slide an ATCA-compliant blade into an ATCA-compliant chassis; you just buy a card with all the standard plug-ins built in,” says Metz.
By changing blades around and sliding in cards or programmable network processors, carriers will be able to jump up the OSI stack and pull whatever information they need for billing.
How It Works
ATCA specifications define measurements and configurations for everything from the chassis, boards, interfaces and management systems, to the processing elements, such as digital signal processors for echo cancellation and high-speed network processors for traffic routing, as well as server processors in enterprise applications.
ATCA standardizes physical form factors describing the size and dimensions of cases, blades, subsidiary cards, the pins and connectors on the back of the board and connectors in the chassis, as well as power sources in the backplane.
That means CIOs can utilize fiberglass with electrical traces rather than multiple electrical wires. With the switch built into the shelf, no cables are necessary to connect and scale servers. “Today, standalone or rackmount servers are interconnected with external switches and cabling, making it cumbersome to build comprehensive server-based billing systems,” says Venkataraman Prasannan, senior director of RadiSys, a provider of embedded building blocks to OEM companies. He contends that ATCA will enable telecom IT organizations to collapse external cabling and switching on to the back plane of the shelf.
ATCA-compliant blade servers address three classes of applications. The first is the control plane, where in the network user data travels and traffic functions take place. Usually a server function, ATCA will open the door for blade servers to handle softswitches and call servers, among other things. Second is the service plane for Web servers and application servers, including billing, database applications and PTT applications. Carriers want highly available applications in order to comply with SLAs, which is why the ability to handle robust applications is an important characteristic of blade servers. Third are management planes, where network functions around network operations centers take place.
“The volumes of ATCA-based equipment will grow
significantly because it has two fabrics—a full mesh and gigabit ethernet,” says Eric Mantion, a senior analyst, networking technologies, with In-Stat/MDR. “At its inception, no one envisioned that each line card would be its own computer,” says Mantion, noting that the back plane on an ATCA full mesh configuration can offer 2.4 terabits per second.
To increase processing under ATCA, CIOs will plug in new blades, which will automatically be recognized as server clusters. “They will no longer have to add and connect servers,” Mantion says.
Whether true plug-and-play will be a reality with ATCA is yet to be seen. “Just because a PC card fits into a modem doesn’t mean it will work. Similarly, a card that slides into a chassis won’t work unless vendors get high-level functionality that works in the same way,” says Mantion.
To allay that fear, participating ATCA companies are participating in “plug-fests,” where every few months, engineers from leading OEMs, computing companies, carriers, service providers and ISPs are tailoring their equipment to comply with ATCA specifications.
Currently, ATCA equipment is prototype gear. In the next year, ATCA chassis and blades will be in trials with production implementations expected by the third quarter of 2006.
Email this article
Add a comment
Printer version
Order reprints
RSS Feed
Bookmark article