Since Mobile World Congress, the telecom industry has amped up the jargon on software-defined networking (SDN) and network functions virtualization (NFV). This is an exciting inflection point for telecom, and an opportunity to continue to deliver on the promise of an open Architecture, ATCA, for equipment used in the telecom market.
However, as the industry goes through the hype curve, I see many citing Amazon, Facebook, Google, and enterprise data-center examples of where the telecom world is heading with NFV — claiming that as that happens, ATCA won’t have a place in the future. I smile and say: Not so fast.
Let me break down why ATCA became the platform of choice for leading telecom and service providers, and why the ATCA ecosystem will continue to evolve and be a significant contributor to SDN and NFV technology advancement.
Myth 1: ATCA is a proprietary and expensive architecture.
Let us first be clear that ATCA is not a proprietary architecture. It is supported by hundreds of companies that offer a wide array of products and the standard is upheld by the industry group PICMG. It offers the ideal business model for business leaders and purchasing teams to drive a highly competitive bidding process. Keep in mind, the solution cost for the network operator is largely a function of the supply-chain models that can be significantly improved even with the same underlying hardware platform — as long as it is open/standards-based. ATCA allows the technology mix to provide best performance/space/Wattage with market-leading technologies on x86s, NPUs, switching and acceleration engines, etc. Proprietary server architectures from single-source vendors cannot make the same claim.
However, ATCA is just a form factor, and when one looks at the needs of SDN and NFV, future hardware platform changes may be required to optimize the performance of the now virtualized software stacks. And that is the strength of an industry-supported, open architecture; the standard can evolve as new requirements emerge. With a proprietary architecture, you must wait for the proprietary box providers to enable the technology.
To the second point — ATCA is not expensive. The key is to determine what you want to do with the platform. If you want a high-availability, NEBS-capable platform with redundancy at the hardware level, there is a “value” delivered for this capability. If you want to have a lower cost-per-board, get the volumes up (a function of supply chain models). A company’s total cost of ownership can be lowered by using commercial ATCA platforms rather than enterprise servers, based on key telecom features requested. These include long life support, redundancy at the hardware layer (not just the software layer), and manufacturing/development services for platform middleware capabilities. Keep in mind that performance and cost will remain the single biggest contributors to capex savings for operators — making a strong case for mixed technologies for years to come for a broad array of telecom applications. ATCA was designed to support these mixed technologies.
Myth 2: Virtualization abstracts the hardware, so it doesn’t matter what platform is used to serve network operator requirements.
Fact: the platform used matters. Here is why.
- Reliability/Availability. the telecom cloud is not the enterprise cloud. A highly regulated environment, such as E911 capability/emergency calls, cannot afford connection drops. Software-based models remain far from proven when it comes to telco-grade reliability and availability. Today’s enterprise data centers handle reliability with software mirroring, and that is great when someone is surfing the web. However, it doesn’t meet the telecom industry’s stringent QoS requirements. Hardware redundancy models delivering five-nines availability (99.999%) are proven and continue to be the strength of the ATCA platform.
- Performance. There are many ways to measure performance. In a telecom environment, latency is critical. While the consumers’ views on acceptable QoS have certainly changed given heavy cell phone use, even mobile operators openly admit that the #1 reason for subscriber churn is poor coverage and QoS. Nobody wants to go back to the age when significant time delays during a phone conversation were the norm. While deployed broadly at Enterprises, virtualization still struggles with significant latency challenges. While there are efforts underway to address this issue, the gap between native and virtualized app performance remains significant. In some applications, it’s acceptable; in others that are data-plane intensive, it will be many years before this issue is resolved.
- Flexibility. Enterprise commercial form factors do not offer the flexibility to mix multiple technologies within the box. This is one of ATCA’s greatest strengths. It offers developers a heterogeneous environment to put the most optimized architecture together to ensure that applications are differentiated from each other’s network elements and also deliver performance/space/Watt/cost advantages. Network processor suppliers and x86 general-purpose suppliers spend billions to ensure usage models run best on their architectures. ATCA remains the open platform architecture of choice for telecom equipment manufacturers, supporting the breadth of NPUs and x86 technologies in the market.
- Space constraints. There are thousands of central offices around the world, with a large majority built in densely populated areas. This is a different model than the mammoth web data centers that are highly centralized. In the telecom industry, space and cost constraints remain a challenge. A highly dense platform like ATCA, which fits in less than a 20-inch depth, can save precious real estate that is needed in these existing central offices. While migration to fewer telecom data centers remains the goal of network operators, it will be a very gradual process and limited when compared to the examples of Amazon, Facebook, and Google data centers. There is just too much risk involved. A decentralized environment for densely populated central offices will remain a regulated requirement for many years. In the meantime, the ATCA ecosystem is well situated with its experience and market penetration to meet today’s space constraints and to evolve the form factor to effectively address the needs of the future telecom infrastructure.
I believe that the telecom network architecture is going through a fundamental shift toward SDN and NFV. As the fundamental network architecture changes, a standards-based solution, like an ATCA platform, will offer a better, more cost-effective, and flexible platform than proprietary solutions. As the market settles out over the next five to 10 years, I expect commercial standards-based platforms, like ATCA, to be the clear choice for SDN and NFV deployment.