The Advanced Telecom Computing Architecture (AdvancedTCA® or ATCA®) is a series of open standard computing platform specifications originally ratified in 2002 to meet the needs of carrier grade communications equipment, primarily for core network telecommunications applications. In the 11 years since ATCA products first appeared on the market, there have been numerous advances in the governing standard, the products available, and the products and applications that are driving its adoption and growth.
The most commonly deployed payload blades are high-performance Intel® Xeon® processor blades, used traditionally for control plane applications, but increasingly used in the packet data path. However, one of the biggest differentiators between ATCA systems and other bladed computing systems is the availability of payload blades with technologies other than Intel Xeon. These include media processing blades with digital signal processors (DSP), and packet processing blades with specialized network processors such as the Cavium™ OCTEON™ II or the Broadcom XLR.
The combination of high-availability architecture, I/O and backplane bandwidth, and a choice of available technologies has made ATCA systems particularly well suited for a large set of telecommunications network applications including radio access network control, core network, and data path equipment.
The unprecedented explosion in network traffic, fueled by smart mobile devices, and online content such as streaming video, has stressed existing telecommunications networks. Carriers have been forced to react by expanding their network capacity.
The fundamental problem facing these carriers, however, is that their revenue is not growing at the same rate. In general, the Average Revenue per User (ARPU) is fairly flat, meaning that, in order to remain profitable, carriers must look for innovative ways to manage their networks other than by simply scaling their existing infrastructure.
ATCA systems are particularly well suited to be used as a platform for such innovative new applications because they uniquely combine the following features:
• High-bandwidth fabrics (40G Ethernet)
• High-performance payload blades optimized for packet processing applications rather than traditional server applications
• High-capacity digital signal processing blades for voice and video optimization.
Two particular areas of focus are network intelligence applications with deep packet inspection (DPI), and mobile data optimization.
In order to remain competitive and profitable, telecommunications carriers cannot be restricted to simply delivering packets over a dumb pipe network. Rather, carriers are looking to introduce systems that gather intelligence regarding the specific nature of the traffic carried by their networks. This intelligence allows carriers to make decisions about delivering that data more efficiently, and allows them to provide differentiated service offerings that improve the experience of users.
Network intelligence applications generally require DPI. In a traditional IP network, network equipment only looks at a packet header, which includes information required to route a packet from the source to the destination. No specific attention is paid to the content of the packet. With DPI, the equipment looks much deeper into the packet, separating the traffic into flows. Each flow represents the packets sent from source to destination for a particular session, and can be further classified by protocol and specific application, often with intermediate tunnels that may or may not be encrypted.
DPI enables a range of network services including network optimization, flow inspection, data flow management, security and application monitoring. These services may be called many things — such as user experience optimization, policy definition and enforcement, quality of service, tiered services, or lawful intercept.
Network intelligence applications are a strong driver for ATCA product innovations today, including driving the introduction of 40G Ethernet fabrics, with associated payload blades based on either traditional packet processors such as the Cavium Octeon II, or increasingly Intel Xeon processors. System software required by such applications include load balancing software that distributes flows evenly across processor arrays, fast path packet processing software, and higher level software stacks.
Voice and Video
Once packet flows are identified, there is an opportunity to manipulate them to optimize content delivery. Video traffic represents a large and growing percentage of network traffic growth, and there is a class of network systems that must manipulate that data in the network to either accommodate different end user devices, or to mix the video streams for applications such as video conferencing or targeted advertising insertion.
Voice traffic is also traditionally manipulated in the network. Various standards exist to encode voice traffic, based on requirements for bit rate, latency and quality. At the border between voice networks, in elements such as session border controllers (SBCs) and media gateways, these voice streams are typically transcoded between these different codecs using specialized hardware employing DSPs.
The need for in-network voice and video transcode functions is another strong driver for ATCA systems due to the availability of specialized DSP blades, employing technologies from companies such as Texas Instruments and Octasic.
Although conceived and developed for the telecommunications market, ATCA systems are beginning to gain acceptance in adjacent markets. The largest of these markets are military, aerospace, and security. All share common requirements for highly reliable, high-performance systems, which are designed to cope with operating environments that are typically harsher than the traditional IT data center.
Data transmission, packet processing and high-performance computing are becoming ever more important functions in military, aerospace and security equipment. This is driving defense and aerospace contractors to examine whether ATCA can be adopted successfully in the tough conditions experienced by soldiers, pilots and sailors.
ATCA meets the US Department of Defense (DoD) mandate for a modular open system approach (MOSA), commercial- off-the-shelf (COTS) and reduced size, weight, power and cost (SWaPC) based solutions.
ATCA for the Cloud
The telecom industry loves to talk about the efficiencies and savings that can be gained by moving services to “the cloud”. But, when people talk about the cloud, they are generally speaking in terms of moving functions somewhere where the responsibility of service delivery becomes somebody else’s problem. The typical view of this cloud is a large array of very inexpensive homogenous servers, with general- purpose server processors, and no application acceleration. Individual nodes would not be highly available. Rather, service availability comes from a very large array of unreliable elements, managed by very smart software. In such a model, is there room for ATCA systems?
When people talk about moving applications and workloads to the cloud they assume that there are nodes in the cloud — that they don’t need to worry about — that will provide 100% reliability and all they have to do is write software that lives in a virtual machine. That ignores entirely the infrastructure to host these virtual machines. So fundamental questions are:
• What are the reliability requirements of these virtualized applications?
• What is the cost of an outage?
These questions are often conveniently ignored. Currently deployed IT virtualization software does not deliver the platform reliability required by telecommunication services. So, the big question for the industry is whether there will continue to be a market for platforms with traditional telecom availability and reliability, which are designed to host these new virtual functions.
This opens the door to telecom-specific virtualization platforms, designed and standardized to run network functions. An industry initiative called Network Function Virtualization is underway with the goal of standardizing these platform software interfaces with telecom-reliable management services.
A second consideration is the performance that can be achieved on general- purpose hardware versus what can be achieved with application-specific hardware. A wise man once told me that you cannot virtualize hardware. There will always be opportunities to optimize hardware platforms to accelerate performance to improve capacity, power efficiency and cost of acquisition and ownership. This application acceleration may take several forms, including specialized ATCA blades, or blades using general-purpose processors that have been optimized for specific applications. Today’s 40G Intel architecture ATCA blades incorporate high-performance I/O, memory and acceleration offload features that are simply not available on standard servers. The same can be said for media processing blades using clusters of desktop processors with GPUs for media processors. Neither can really be called traditional cloud elements.
So the future of ATCA may be cloudy, but the need for telecom-specific platforms with features such as high availability, high-bandwidth fabrics, and application-specific acceleration will ensure that demand for such open standards- based platforms remains strong for many years to come.
What’s next for ATCA?
Future investment in ATCA will likely be focused in two areas:
• More power and cooling
• Higher bandwidth fabrics
The ATCA standard was originally conceived to accommodate payload blades that nominally consume up to 200W of input power and cooled by approximately 40 CFM of airflow per slot. However, new technologies have resulted in blades that have pushed through that envelope, and the standard is being accordingly evolved to approximately double those numbers. This should result in enough power and system airflow to accommodate several technology insertion cycles.
For the ATCA fabric, a logical next step would be the evolution of the current 40G Ethernet fabric to 100G. This step will require first the standardization of 100G Ethernet over a copper backplane by the IEEE, followed by the adoption of a 100G backplane standard by PICMG for ATCA. Such work is currently underway, and will most likely be complete within a few years.
Demand for ATCA systems is strong and is expected to continue to grow. Within the telecommunications market, demand for network intelligence and mobile data optimization services will drive demand for open standards based systems with high performance fabrics and application-specific acceleration. Outside the traditional telecommunications market, relevance and demand is growing in markets such as military and aerospace.
This article was written by Robert Pettigrew, Marketing Director, Embedded Computing, Emerson Network Power (New York, NY). For more information, visit http://info.hotims.com/45606-400.