Skip navigation
2012

Graphics processors that are highly optimised for graphics processing are also very efficient for other media processing functions such as video transcoding. The latest processors for mobile, desktop and workstation platforms integrate high-performance graphic cores opening up new opportunities for systems with integrated media processing and other functions such as cryptography or control plane processing.

 

In this blog I am going to explore the benefits of using 3rd generation Intel® Core TM processors, Ivy Bridge, for video transcoding and other related applications. For this blog I have been talking to Paul Stevens, Telecom Sector Marketing Director for Advantech in Europe.  Advantech is premier member of the Intel® Intelligent Systems Alliance. The 200-plus members of the Alliance collaborate closely with Intel® to create hardware, software, tools, and services to help speed intelligent systems to market.

 

Video Transcoding

 

Video transcoding is a key function required to provide video services to mobile and other network users allowing operators to add subscriber–specific logos and text overlays, advertisements and interactive user applications. The percentage of mobile traffic used by video services has been steadily increasing. According to the Cisco Visual Networking Index (VNI) Global Mobile Data Traffic Forecast Update from February 2012, mobile video will increase 25-fold between 2011 and 2016 and will account for over 70 percent of total mobile data traffic by 2016.

Video transcoding has been traditionally implemented using dedicated hardware such as application-specific integrated circuits (ASIC) or digital signal processors (DSP). The latest systems need to support a wide range of media processing and other functions. This workload consolidation requires significantly more flexible solutions. The combination of general purpose Intel® architecture (IA) cores and integrated graphics processors that can be programmed for video transcoding and other media processing is both flexible and scalable.

 

Ivy Bridge Benefits        

 

The 3rd generation Intel® Core TM processors, Ivy Bridge, integrate 2-8 IA processor cores with Intel® Hyper-Threading technology. Key benefits of most 3rd generation Intel® Core TM processors include Intel® Turbo boost technology and performance tuning, delivering high performance and low power consumption. Most 3rd generation Intel® Core TM processors also integrate AES instructions for security processing and Intel® HD Graphics 4000 with DirectX 11 support.

 

The Intel® HD Graphics 4000 on 3rd generation Intel® CoreTM processors is twice as fast as Intel® HD Graphics on 2nd generation Intel® Core™ processors. This processing power is also available for video transcoding and other media processing applications in systems where graphics is not required. Media processing applications can be optimised using the Intel® Media Software Development Kit (Intel® Media SDK 2012).

 

Intel® Media SDK 2012 

 

The Intel® media SDK 2012 is a cross-platform application programming interface (AP) for developing media applications. The Intel® media SDK 2012 supports the hardware accelerated video encoding, decoding and transcoding provided by the Intel® HD graphics 4000 integrated into the third generation Intel® CoreTM processors. The Intel® media SDK 2012 significantly reduces the development time for media applications that take advantage of the underlying hardware performance. The Intel® media SDK 2012 supports a range of Codecs and features including H.264, MPEG-2 and MVC video encoders and decoders and various video processing filters.

 

AMC and MicroTCA       

 

The AdvancedMC (AMC) module is a flexible form factor for a wide range of embedded applications including telecom. AMC modules can be used in proprietary systems or standard platforms such as MicroTCA and AdvancedTCA (ATCA).

 

MIC-5603.jpg

Figure 1. Advantech MIC-5603 Processor AMC

 

The Advantech MIC-5603 Processor AMC module is shown in Figure 1 is based on the 3rd generation Intel® Core TM i7 processors. The module has an optional front panel HDMI port that is connected to the processors integrated Intel® HD 4000 graphics core. The solution supports DirectX v11, OpenGL v3.1 and OpenCL v1.1. The Advantech MIC-5603 Processor AMC module can be programmed for transcoding and other media applications using the Intel® media SDK 2012 and readily available software.

 

Larger systems can be built using a MicroTCA chassis as shown in Figure 2. A single MicroTCA chassis will support up to 12 AMC modules.

 

MicroTCA Chassis.jpg

Figure 2. MicroTCA Chassis

 

Integrated Graphics Enables New Applications

 

The integration of Intel® HD Graphics 4000 into the 3rd generation Intel® Core TM processors opens up new opportunities for system developers. By using processor AMCs , such as the Advantech MIC-5603 Processor AMC module, and the Intel® media SDK 2012, system developers can quickly integrate video transcoding capabilities into mobile network systems and other solutions. The high level of video processing performance enables ‘transcoding on the fly’ of multiple HD video streams per processor. This ability to transcode, reformat, modify and manipulate video streams within the mobile network infrastructure is the foundation for new, value added services allowing the operator to develop applications that combines subscriber, and network and location information to enhance the user’s video experience.

 

security_analytics.pngTo learn more about advanced signal processing, see intel.com/go/embedded-analytics

 

comm.pngFor more on flexible, scalable, standards-based communications visit intel.com/go/embedded-communications

 

Advantech is a premier member of the Intel® Intelligent Systems Alliance.

 

Simon Stanley

Roving Reporter (Intel® Contractor), Intel® Embedded Alliance

Principal Consultant, Earlswood Marketing

Follow me on Twitter: @simon_stanley

AdvancedTCA (ATCA) based systems are reaching new performance levels with the latest Intel® Xeon® Processors E5-2600 and E5-2400 Series. These processors provide several key features that increase ATCA platform throughput including more cores, enhanced power management, and greater memory and I/O bandwidth.

 

In this blog I am going to explore the latest developments in the ATCA market and the benefits that Intel® Xeon® Processors E5-2600 and E5-2400 Series bring to these platforms. Many of the 200-plus member companies in the Intel® Intelligent Systems Alliance provide solutions for ATCA. For this blog I have been talking to four members of the alliance that have both processor blades and complete application ready ATCA platforms; Advantech, Emerson, Kontron and ADLINK Technology.

 

ATCA and the Shift to 40G          

 

ATCA was designed as a common platform for telecom systems. ATCA chassis have 2-16 slots with shared power supplies, cooling units and management cards. Each slot can have a large front blade and a much smaller rear transition module (RTM). The front blades can be any mix of processor, storage and I/O blades. Larger systems use 1, 2 or 4 slots for switch blades. The most common configuration is a 16 slot system with two switch blades providing redundancy.

 

40G ATCA Platform.jpg

Figure 1. 40G ATCA Platform

 

First and second generation ATCA-based systems had Gigabit Ethernet (GE) switch blades. Third generation ATCA systems with 10GE switch blades started shipping in 2007 and  forth generation systems with 40GE switch blades became available in volume at the start of 2012. ATCA chassis and backplanes and have also developed to support the higher speed interconnects and more power per slot, beyond the original 200W specification. The latest ATCA chassis will support 40Gbit/s interconnects and 350W or more per slot. This is opening up new applications for ATCA within and beyond telecom.

 

ATCA Expands Beyond Telecom               

 

ATCA is already widely used within telecom with most tier 1 vendors and many tier 2/3 vendors building ATCA-based systems. Most first and second generation ATCA-based systems were used in wireless infrastructure for 3G networks. Third and fourth generation ATCA-based systems are used by several leading vendors for the enhanced packet core (EPC) and other parts of the 4G/LTE wireless infrastructure including policy management.

 

10G and 40G ATCA-based systems are now used in a wide range of applications beyond wireless infrastructure including security, IPTV and wideband access controllers. ATCA is also being used in enterprise networks, scientific research and military applications.

 

Technology Trends        

 

Over the past few years there have been some significant changes in the way companies are using ATCA. Many companies building ATCA-based systems have integrated chassis, blades and platform software in-house. There is now a noticeable trend away from in-house integration towards sourcing application ready platforms. This change is largely driven by the cost of integrating new 10G and 40G systems and the ready availability of 10G and 40G application ready platforms with the latest technology including the Intel® Xeon® Processors E5-2600 and E5-2400 Series.

 

The performance increase delivered by these latest solutions is enabling companies to use a single platform across a wide range of applications. In many cases companies can support several applications using a single processor and switch combination, scaling the number of processor blades to meet the specific requirements. This trend is driving an increase in the use of Intel architecture (IA) processors with companies using IA processors for both control and data plane applications.

 

The use of IA processors for data plane applications has been made significantly easier with the introduction of the Intel data plane development kit (DPDK). The DPDK is a set of source code programming libraries that implement basic packet processing functions such as moving data in and out of Intel processors. Many companies developing ATCA-based systems are already taking advantage of the DPDK for data plane functions such as deep packet inspection (DPI) for policy management, security and other applications.

 

Enabling Performance

 

Intel® Xeon® Processors E5-2600 and E5-2400 Series have up to 8 cores, four DDR3-1600 memory controllers, two 8GT/s QPI interfaces and 20MB layer 3 caches. This enables significantly higher performance than previous generations. The Intel® Xeon® Processors E5-2600 and E5-2400 Series also have up to 40 PCI Express 3.0 lanes supporting I/O bandwidth significantly in excess of 100 Gbps. For ATCA this allows support for dual 40GE network interfaces and additional PCI Express lanes to support network interfaces on the front panel and storage or further interface options on the rear transition module (RTM).

 

40G ATCA Processor Blade.jpg

Figure 2. 40G ATCA Processor Blade with Intel® Xeon® Processors E5-2600 Series

 

 

Multivendor Ecosystem

 

The ATCA market is now supported by a mature ecosystem. ATCA processor blades integrating Intel® Xeon® Processors E5-2600 and E5-2400 Series are available from more than four vendors giving system developers significant choice. By using these blades equipment manufacturers can quickly upgrade their ATCA systems to take advantage of the performance uplift from Intel® Xeon® Processors E5-2600 and E5-2400 Series or introduce new systems based on 40G ATCA platforms that will give them market leading solutions.

 

To learn more about achieving industry-leading performance, see embedded.communities.intel.com/docs/DOC-7419.

 

comm.pngFor more on flexible, scalable, standards-based communications visit intel.com/go/embedded-communications

 

Advantech, Emerson and Kontron are premier members of the Intel® Intelligent Systems Alliance. ADLINK is an associate member of the Intel® Intelligent Systems Alliance.

 

Simon Stanley

Roving Reporter (Intel Contractor), Intel® Embedded Alliance

Principal Consultant, Earlswood Marketing

Follow me on Twitter: @simon_stanley

By using Intel® Xeon® processors E5-2400/2600 (formerly codenamed “Romley”) system developers can build powerful server platforms with up to 68 processor cores for every 1U of rack space. This allows companies to meet the demand for more data intensive computing solutions within current space and power consumption restrictions.

 

In this blog I am going to explore the different approaches to packing high-performance processors into limited rack space and the benefits that Intel® Xeon® processors E5-2400/2600 bring to server platforms. A number of companies in the Intel® Intelligent Systems Alliance provide market-ready server solutions. For this blog I have been talking to Austin Hipes, VP of Technology at NEI, an affiliate member of the Intel® Intelligent Systems Alliance.

 

Cloud and Traditional Data Centers Driving Server Densities

 

The demand for higher density server solutions continues to grow, driven by the requirements for cloud and traditional data centers, and high-performance computing (HPC). Cloud data centers include large numbers of servers configured as virtual machines serving many customers. Traditional data centers range from small units with a few servers up to large server farms run by international organizations, such as Amazon and Google, and financial institutions.  HPC sites also require large numbers of servers.

 

1U/2U servers have traditionally been used in small and many larger data centres. Each server is self-contained with motherboard, storage, power supplies and cooling. This approach, that has been very successful in providing stable platforms and a simple architecture, no longer delivers the density required for larger data centers that must fit within tight space and power consumption constraints.. Even with the latest Intel® Xeon® processors E5-2400/2600 the maximum number of cores per 1U of rack space using traditional 1U/2U dual processor servers is 16.

 

Multi-Node Servers

 

Multi-node servers pack four servers into a single 2U chassis. These servers are built using smaller motherboards with a reduced number of PCI Express slots and fewer memory modules than a standard server. Each server has a dedicated disk and I/O interfaces but  shares power and cooling infrastructure with 1-3 other nodes. Multi-node servers with Intel® Xeon® processors E5-2400/2600 are available from a number of well-known vendors including Dell, Fujitsu, HP and QuantaQCT. A quad-node rack mount server is shown in Figure 1. This approach supports up to 32 processor cores per 1U of rack space.

 

Intel H2000 quad-node server.png

Figure 1. Quad-node rackmount server

 

Blade Servers

 

Blade servers have been widely deployed and consist of a 6U, 7U, 9U or 10U chassis with up to 20 server blades in each chassis. Most blade servers have redundant power supply and cooling units built into a proprietary chassis. I/O from each server blade is switched through switch modules also plugged into the chassis. Blade servers support advanced system management and hot swapping of blades and other components. Blade servers are available from several vendors including Dell, HP, IBM and Supermicro. Blade servers will support up to 240 Intel® Xeon® processors E5-2400/2600 in 42U rack. This provides a density of 45.7 processor cores per 1U of rack space.

                                                                                   

“Bare Bones” Servers

“Barebones” servers combine many benefits of both multimode servers and blade servers. “Bare bones” servers are built using stripped down motherboards with only the features required for servers. A multi-node server blade is shown in Figure 2. System management is kept to a minimum reducing cost and complexity. This approach allows higher density systems using off the shelf hardware as promoted by Facebook through the Open Compute Project.

 

High Density Customized SBC.png

Figure 2. Multi-node server blade

 

“Bare bones” blade servers are ideal for large data centers. These blade server systems with that Intel® Xeon® processors E5-2400/2600 blades and back to back stacking can have 96 blade slots in a 45U equivalent rack as shown in Figure 3. Bottom to top cooling significantly increases the total power envelope for the system. Each 550W server slot can support up to four 8 core CPUs giving a total of 3072 cores and a density of 68 cores per 1U of rack space.

 

High Density back-to-back SBC rack.png

Figure 3. High Density “Bare Bones” Blade Server

 

Benefits from using Intel® Xeon® processors E5-2400/2600

The Intel® Xeon® processors E5-2400/2600 integrate up to 8 cores per socket, 33% more than the previous generation devices. The processors integrate advanced power management and are available with power ratings from 50W to 135W. The devices support three or four low voltage DDR3 memory channels delivering higher performance and lower power. The integration of PCI express 3.0 interfaces reduces the number of I/O cards needed.

 

The combination of more cores and higher integration makes the Intel® Xeon® processors E5-2400/2600 ideal for high density server systems in cloud and traditional data centers, and high-performance computing. The enhanced power management and low power options enable server densities up to 68 cores per 1U of rack space.

 

efficiency.pngTo learn more about power-efficient performance, see intel.com/go/embedded-energyefficiency.

 

comm.pngFor more on flexible, scalable, standards-based communications visit intel.com/go/embedded-communications

 

Dell and HP are associate members of the Intel® Intelligent Systems Alliance. NEI and SuperMicro are affiliate members of the Intel® Intelligent Systems Alliance.

 

Simon Stanley

Roving Reporter (Intel Contractor), Intel® Embedded Alliance

Principal Consultant, Earlswood Marketing

Follow me on Twitter:@simon_stanley

Wire-speed DPI and other packet processing functions can be now be implemented without developing specialized packet processing solutions. Companies can easily develop systems that will support packet processing at 40Gbit/s and above by using the Intel® data plane development kit (DPDK) and Intel® Xeon® processor E5-2400/2600 (formerly codenamed “Romley”) together with off-the-shelf software and hardware solutions.

 

In this blog I am going to explore the benefits of using the standard Intel® Xeon® processors for packet processing and the support that is available from Intel and companies in the Intel® Intelligent Systems Alliance to significantly reduce time to market. The 200-plus members of the Alliance collaborate closely with Intel to create hardware, software, tools, and services to help speed intelligent systems to market. For this blog I am using examples from two members of the Alliance; Radisys and Wind River that have been collaborating closely with Intel on the development and use of the DPDK.

 

Multicore Packet Processing

 

High performance packet processing (data plane) has historically been handled by dedicated hardware or network processors. These devices must be closely coupled to a suitable general-purpose processor to handle control packet processing (control plane) and system management. This leads to an inflexible system in which it is not easy to balance the processing capabilities for data plane and control plane. Similar problems balancing the work load can also occur with multicore processors that integrate both general-purpose processor cores and hardware acceleration for packet processing.

 

The Intel® Xeon® processors E5-2400/2600 have several features that make these processors suitable for DPI and other packet processing applications. These processors integrate up to 8 Intel Architecture cores with large layer 2 and layer 3 caches, and high-performance DDR3 memory controllers, delivering very fast memory access, critical for DPI. The Intel® Xeon® processors E5-2400/2600 have integrated SIMD (Single-Instruction, Multiple-Data) cores providing a very powerful parallel processing capability and support instruction extensions for application-specific functions such as security.

The flexible power management on the Intel® Xeon® processors E5-2400/2600 allows individual cores to be slowed down, minimizing power consumption. This is particularly important when cores are allocated to completely different applications such as packet processing and system management. These processors also integrate up to 40 PCI Express Gen 3 interfaces supporting a total I/O bandwidth significantly in excess of 100Gbit/s.

 

Benefits from Using DPDK

 

The Intel® data plane development kit (DPDK) is a set of source code programming libraries. These are small algorithms that implement basic packet processing functions such as moving data in and out of Intel processors. These algorithms make maximum use of the layer 2 and layer 3 caches and the direct memory access (DMA) engines in the memory control blocks. The libraries are optimized for the latest Intel® processors.

 

By using the DPDK system developers can quickly implement packet processing functions on Intel processors such as the Intel® Xeon® processors E5-2400/2600. This approach has already proved very successful with many system developers using the DPDK to develop packet processing solutions for applications such as policy management.

 

Integrating DPDK with Hardware Platform

 

The Intel DPDK programming libraries have been developed to optimize system performance using the standard features of a Intel® Xeon® processors E5-2400/2600 based system. Most embedded and networking systems use additional hardware such as network interface cards (NIC) or Ethernet controller devices. These NICs and Ethernet controller devices integrate dedicated hardware to accelerate packet transfer to and from the system processors. To gain maximum performance with DPDK system developers need to port the drivers for these NICs and Ethernet controller devices. This can be a significant challenge.

 

Fortunately Intel® Intelligent Systems Alliance members are already porting drivers for various NICs and other hardware. An example is the Radisys ATCA-4600 ATCA processor blade (see Figure 1). This integrates dual sockets for E5-2448L or E5-2428L Intel® Xeon® processors and a Mellanox Ethernet controller device supporting dual 10 Gbit/s or 40 Gbit/s interfaces. For this blade Radisys has ported the Mellanox Ethernet controller driver to work with the Intel DPDK giving customers a fast time to market solution to implementing line rate packet processing.

 

Figure 1.jpg

Figure 1. Radisys ATCA-4600 with dual socket E5-2448L or E5-2428L Intel® Xeon® processors

 

Radisys is working closely with Intel and expects to have both ATCA blades and updated drivers available for future Intel Xeon processors. This is a major benefit to system developers allowing a quick upgrade to the latest processors without sacrificing driver efficiency.

 

Integrating DPDK with Operating System

 

The DPDK libraries are not targeted at a specific operating system (OS) or hardware platform. To make maximum use of the DPDK system developers need to integrate the DPDK algorithms with the operating system and packet processing implementation they are using. This integration can also be a significant challenge as it is important to optimize packet processing performance.

 

Intel® Intelligent Systems Alliance associate member Wind River will provide an integrated, embedded development kit for next-generation servers, data centers, and storage and communication systems. This includes Wind River Linux, to meet Carrier Grade Linux 5.0 specification, and integrated Intel DPDK libraries. The company also has a network acceleration platform that integrates a lightweight hypervisor and accelerated IP stacks. The Wind River Network Acceleration Platform uses the DPDK libraries for accelerated packet delivery . These solutions significantly reduce time to market for high-performance carrier-grade systems.

 

Growing Ecosystem

 

The introduction of Intel architecture multicore processors with integrated high-performance I/O and memory controllers like the Intel® Xeon® processors E5-2400/2600 has opened up new opportunities for high-performance packet processing without using specialized hardware. The Intel DPDK and off-the-shelf hardware and software solutions from Intel® Intelligent Systems Alliance members, including Radisys and Wind River, has significantly reduced the time to market for systems implementing high-performance packet processing such as DPI. Together with Intel these companies are developing a growing ecosystem around the Intel DPDK with hardware specific drivers, support for standard operating systems and timely upgrades to support the latest Intel processors.

 

Icon 1.png For more on building flexible networking solutions, see intel.com/go/embedded-consolidation.

 

Icon 2.png For more on flexible, scalable, standards-based communications visit intel.com/go/embedded-communications.

 

Radisys is a premier member of Intel® Intelligent Systems Alliance. Wind River is an associate member of the Intel® Intelligent Systems Alliance.

 

Simon Stanley

Roving Reporter (Intel Contractor), Intel® Embedded Alliance

Principal Consultant, Earlswood Marketing

Follow me on Twitter: @simon_stanley

Filter Blog

By date: By tag: