AdvancedTCA Newsletter |
Vol. 7, No. 3 June - July 2011 Newsletter circulation: 13,142 |
| www.atcanewsletter.com | |
In This Issue
Announcement Tail-f Systems announced its first annual conference for developers. Tail-f Developer Days 2011 will provide Tail-f Systems users two days of in-depth technical sessions and hands-on labs. Held in Silicon Valley, California on September 22nd and 23rd, the event will bring together developers and members of Tail-f Systems' core technical teams for training and workshops on developing network management applications that leverage the advantages of Tail-f Systems' ConfD and NCS products. Tail-f Developer Days 2011 is open to Tail-f Systems' customers and others by direct invitation. To find out more and register for this event, email training@tail-f.com. Table of Contents
2. Industry NewsMarket Research Crystal Cube Consulting (CCC) has released its report COM Express: Leading the Small Form Factor Board Race. The report covers five year forecasting on COM Express and xTCA platforms (AdvancedTCA, MicroTCA and AdvancedMC). Among the significant findings was 120% sales growth in COM Express from 2009 to year end 2010, associated with over $130 million in sales. With typical module prices in volume of $500 to $1,000, this corresponds to 130,000 to 260,000 modules shipped. The large number of applications for COM Express and their significant volumes (often in the thousands or tens of thousands) indicates that such growth should continue for several years. To view an abstract of the report, visit: www.crystalcubeconsulting.com/English/Reports/COMExpress_Leads_Small_Form_Factor_Board_Race.html Hardware and Software Emerson Network Power announced availability of its first 40G AdvancedTCA payload blade, the ATCA-9405. The ATCA-9405 can provide telecom equipment manufacturers with up to four times performance increase over previous generation ATCA blades for high touch and high throughput packet processing applications in the latest data-intensive network evolution, including:
Featuring dual OCTEON II CN6880 processors from Cavium, each with 32 cnMIPS cores running at up to 1.5GHz, the Emerson Network Power ATCA-9405 enables the implementation of packet processing applications at line speeds of 40Gbps in conjunction with Emerson’s range of 40G ATCA platforms. The new blade supports a variety of software, including Wind River’s Linux 4.x operating system, Cavium’ s packet processing software development kit, 6WIND’s 6WINDGate fast path networking stacks, and field proven L2 and optional L3 Ethernet switch management. More information about ATCA products and technology from Emerson Network Power can be found at www.Emerson.com/ATCA. PT announced the development of a new open-standards solution set, code named Monterey. Monterey is an enhanced application-ready system that combines compatibility with the latest MicroTCA.4 standard with superset extensions including increased backplane performance and higher available power budget. The Monterey architecture leverages the latest trends in high-density computing and high speed fabrics in a compelling cost-performance, space-efficient model that is ideally suited as the foundation for a wide range of next-generation network communications and services applications. It is smaller in scale than ATCA, allowing developers to right-size the solution to best fit their application, thus eliminating the development and support costs of integrating disparate functions onto a single ATCA blade. In addition to the new MicroTCA.4 double-wide form factor, Monterey also supports single width Advanced Mezzanine Card (AMC) modules, leveraging the large installed base of existing AMC technology. PT projects Monterey to be commercially available in September 2011. For more information, visit http://www.pt.com/page/monterey. Elma Bustronic has introduced its 40G AdvancedTCA backplanes, which are based on the design principles of IEEE 802.3-2008 and IEEE 802.3ba-2010 standards. The 40G ATCA backplanes (4 channels x 10Gb/s) feature a controlled-impedance stripline design of 18 layers in a Dual Star routing configuration. The backplane results were achieved using extensive pre-layout and post-layout simulation studies. Signal integrity simulations performed at Elma Bustronic have demonstrated that 10Gbit/s/link performance can be achieved with reasonable costs and manufacturing processes. For superior performance, Nelco 4000 13-SI high grade laminate material was used and the backplane was backdrilled to minimise stub reflections. Other features of the 40G ATCA Backplane include dual shelf manager connectors in slot 0 for redundant IPMB implementation and up to 400W/slot 48VDC distribution to each slot. The shelf managers on the left side of the backplane can be front-pluggable cards in slot 0 while allowing a full 14 slots. For more information, visit www.elma.com. Adax, SURF Communication Solutions, and RadiSys announced the successful delivery of a key component in the migration to LTE – CSFB (Circuit Switched Fallback) – to a major APAC provider of integrated wireless solutions. CSFB enables mobile operators to quickly and economically roll out voice services in conjunction with their LTE networks. It allows mobile devices to “fall back” to GSM or UMTS domains for incoming or outgoing voice calls. Subscribers maintain access to the wide array of rich circuit-switched capabilities, including international roaming while enjoying LTE broadband access to the Internet. The joint solution is NEBS-ready and is based on cutting-edge commercial off-the-shelf technologies from all three partners, integrating Adax’ flexible HDC3 interface cards and SURF's voice processing subsystem into RadiSys’ ATCA server platform. The joint CSFB solution was designed for existing TDM-based network services. RadiSys’ carrier-grade TCA-6006 5U platform, equipped with SBCs and carrier blades, hosts the Adax HDC3 and SURF/Rider AMC cards. Positronic announced that its SAE AS9100 certified manufacturing facility in Ponce, Puerto Rico has achieved Rev C status under AS9100. SAE AS9100 specifications require an organization’s quality management system to demonstrate the ability to consistently provide product that meets customer and applicable statutory regulatory requirements. This goal is accomplished through the effective application of the quality management system, including processes for continual improvement of the system. 3. Navigating the Technology Minefield, from Lance Leventhal, Technology EditorLooking Forward at 100G “Ever forward but slowly…” – von Blucher (leading the Russians against Napoleon) As systems move up in frequency, signals attenuate faster, leading to more errors and generally poorer system performance. One way of compensating is to use forward error correction (FEC). However, implementing it efficiently is difficult and requires an understanding of complex mathematical techniques. If you need an immediate solution, you should consider the IP cores developed by Altera for their Stratix FPGAs. These ultra high gain, hard decision EFEC (enhanced FEC) cores offer low overhead, increased transmission distance, and lower transmission power. They can help you solve the problems of 10G, 40G, and even 100G systems. For more information, see www.altera.com. The author would like to thank Alfred Whiffen of Altera’s Newfoundland Technology Center (located in sunny Mount Pearl, Newfoundland) for explaining the concept to him at a recent OFC/NFOEC Conference. Face Up to Facebook…and Twitter ”Nun Banished From Order for Spending Too Much Time on Facebook.” – Atlantic headline, February 2011 “If you're finding it difficult to spend an hour of your waking life without checking or thinking about Facebook, you may be looking for a way out.” – WikiHow, April 2011 Are you on Facebook and Twitter? Is your organization? Nowadays a presence on both sites is as essential as having a telephone number, a fax number, or a Website. But how do you use them effectively? Do they have a role within your organization as well as outside? What about similar Websites that might be useful, including ones aimed at the enterprise? How about security and usage policies? Obviously, both issues could open up huge problems, risks, and liabilities. You may need training in the area – sounds strange, but you have to keep up or else suddenly you’re a dinosaur (imagine someone who doesn’t have a Website or a cellphone). An outfit with the great-sounding name of Ultimate Knowledge Institute provides both training and certification in social media technologies. They offer classes for practitioners, engineers, security professionals, and managers. They are, of course, on the Web at www.ultimateknowledge.com and presumably available also via Facebook and Twitter. I would like to thank their VP Business Development, William Rybczynski, for explaining what they do at a recent C4ISR event. Making Multicore Compute For You “A million monkeys typing still won’t produce Shakespeare.” – anonymous saying One problem with most multicores is that the cores usually aren’t very powerful. In order to fit lots of them on a single chip, designers have made them quite simple. The end result works well if such simple cores can handle your problem. But what if it requires a lot of floating point or other computations? Then you may want to consider a new multicore offered by a startup called Adapteva. Their Epiphany product line has up to 4,096 processors on a chip, each of which is a fully-featured floating point RISC processor. It also offers low power, scalability, ease of integration, and ease of use via standard C programming. You won’t need expensive ASICs or FPGAs to handle high-computation workloads in small mobile devices. One can imagine UAVs, medical instruments, space systems, vehicles, and satellites as among the applications. For more information, see www.adapteva.com. The author would like to thank Adapteva CEO Andreas Olofsson for explaining the concept at a recent Embedded Systems Conference. Get Social…In the Enterprise “Man is a social animal.” – Spinoza Facebook is a great way to communicate with a lot of people, even if you don’t know their email addresses. It’s also a great place to post news and pictures, ask questions, and share information. But obviously Facebook is an open world with no secrets (as several public figures have discovered recently) and limited security. Wouldn’t it be nice to have a social network that extended only across your organization and was enterprise-strength? Well, such a network exists under the odd name of Yammer. Yammer allows everyone in your organization to collaborate and communicate. As with Facebook, you can post information, upload pictures, have private dialogs, and share documents. And, of course, it doesn’t matter whether the other users are next door, halfway around the world, or somewhere in-between. Nor does it matter whether people are on a desktop,notebook, netbook, tablet, or mobile phone. And communicators aren’t filling up everyone’s email in-boxes with lots of unwanted copies. Did I mention that Yammer is free? Well, it is. Just go to www.yammer.com to sign up. As you might expect, there’s a premium version with added features for a price. The Website has testimonials, and Kathleen King of Alabama Power has provided an unsolicited one in a recent issue of CIO Magazine. Well worth investigating! 4. Organization UpdatesAdvancedTCA and AXIe – Serving Both Markets Many observers have noted the differences between AdvancedTCA and AXIe, a derivative for the test and measurement community. However, although AXIe backplanes add instrumentation specific topologies (either to Zone 2 for AXIe 1.0 or Zone 3 for AXIe 3.1), there is no reason why processing blades cannot be both AdvancedTCA and AXIe compatible simultaneously, thus serving both markets. AXIe does not disturb AdvancedTCA’s basic communication structures. The base interface remains Ethernet, and the fabric can be PCIe, as specified in PICMG 3.4. The special pins that enable hardware triggers and timing are not required, and typically not needed in a processing blade. For example, the Agilent M9536A AXIe Embedded Controller is actually a 100% AdvancedTCA compatible blade. Its fabric supports PCIe, the defacto communication standard for AXIe. Although PICMG 3.4 specifies only PCIe Gen 1 capability, the M9536A supports Gen 1 and Gen 2 speeds, but trains down to Gen 1 speeds in a Gen 1 system. AXIe backplanes differ from AdvancedTCA in that slot 1 also is a hub for specialized triggering fabrics, hosting the ASM (AXIe System Module), which provides the switching for the data fabrics and AXIe triggering fabrics. However, AXIe processor blades such as the M9536A don’t require fabric hub capability, since they serve only in node slots, and have no need to add the specialized AXIe triggering functions. This is why a fully AdvancedTCA-compatible blade can be deployed in AXIe node slots. The M9536A is designed for measurement applications in electronic test, communications, and high-energy physics. It features an Intel Xeon L5518 quad-core processor running at 2.13 GHz, 8 GB of RAM, and an on-board 160GB solid-state drive. There are no Zone 3 connections. It can thus operate in AXIe 3.1 systems in which Zone 3 is already defined, or in any slot without matching to a specific Rear Transition Module. Periodic user re-configurability is common in instrument systems, so removing the dependence on an RTM is advantageous. Many instrument systems must control USB and LAN-based instruments. Therefore, the M9536A has three USB slots and two LAN interfaces on its front panel. GP-IB instruments may be controlled using an external LAN to GP-IB converter. A front panel video port supports UXGA resolutions up to 1600 x 1200. It supports either Microsoft Windows or Linux. What does this all mean? Good news for AdvancedTCA vendors! The M9536A proves that a product that is essentially an AdvancedTCA processor blade can readily address the AXIe marketplace. The most important requirement is to support PCIe communications, which has already been specified in PICMG 3.4. By paying attention to a few details, AdvancedTCA vendors can leverage their product development into a new and growing market without sacrificing the interoperability that is essential in telecom and military applications. Larry Desjardin in Chairman of the Board of Directors of the AXIe Consortium, and General Manager of the Modular Product Operation at Agilent Technologies. You can reach him at larry_desjardin@agilent.com. Cloud Computing for Telecom and Real-Time Services Given the growing popularity of cloud computing (Gartner recently identified it as a top technology trend for 2011), most major IT companies and many telecommunication providers have started to roll out offerings at all service levels: Infrastructure as a Service (IaaS), Platform as a Service (PaaS), and Software as a Service (SaaS). SCOPE’s new white paper “Telecom Grade Cloud Computing”, available at www.scope-alliance.org, describes how to make cloud computing usable for telecom and real-time services. It includes the role and importance of inter-cloud architectures as well as the usage of private, public, and hybrid solutions for telecom and real-time services. SCOPE’s mission is to promote a vibrant and diverse ecosystem of COTS carrier-grade platform components utilizing open standards. This is in line with the cloud computing goal of reducing platform costs while continuing in the direction of more open architectures. “Telecom Grade Cloud Computing” offers a telecom perspective on cloud computing standardization efforts and a common work agenda for SCOPE in its relationship with standardization bodies. It is intended for:
Check out “Telecom Grade Cloud Computing” to learn about how to make cloud computing usable for telecom and real-time services. For more information about the SCOPE Alliance, please visit www.scope-alliance.org. Timo Jokiaho, CTO Base Platform at Huawei Technologies, is the 2011 Chair of the SCOPE Alliance. You can reach him via the Alliance’s press representative, Karen Riley Sawyer of Nereus, at krileysawyer@nereus-worldwide.com. Cloud Outages Raise Availability Concerns Cloud computing is currently a major topic in information technology. It allows organizations to reduce IT costs (and avoid major technology headaches) by not having to manage their own storage, servers, or databases. The arguments in its favor are so compelling that, according to Gartner, cloud computing services will grow to a more than $112 billion industry by 2015. However, highly publicized cloud outages in the news recently, such as ones at Amazon and Microsoft, have raised concerns and made organizations wary about moving essential applications to the cloud. Such outages are obviously service availability failures, and, as we well know, designers can avoid them by planning for high availability ahead of time. Cloud Outages Prove Costly Service Availability in the Cloud Service Availability Specifications John Fryer is President of the Service Availability Forum. He is also Vice President, Business Development, at GoAhead Software. You can reach him at JFryer@goahead.com. OpenSAF Announces First European Symposium Following the success of the 2011 OpenSAF Conference and the positive feedback from the high availability community, OpenSAF is pleased to announce it will host its first one-day symposium on September 13, 2011 in Stockholm, Sweden. It will feature sessions led by well-known technologists, architects, and academics in the HA ecosystem. Key sessions include:
If you are interested in attending the symposium, please register using the short form. To view audio and visual presentations from the 2011 OpenSAF Conference, click here. Monica Hatlen is the President of OpenSAF and Product Manager for Platform and Component Strategy at Ericsson. You can reach her at monica.hatlen@ericsson.com. Higher Port Speeds for RapidIO A new RapidIO roadmap for a serial 10xN set of specifications advances the protocol to lane speeds of 10 Gbps and higher, supporting individual port speeds that scale beyond 100 Gbps. The roadmap satisfies the needs of RapidIO customers in the wireless, defense, aerospace, imaging, video, and server markets beyond what they can implement with other interconnect protocols. It also sets the stage for OEMs to develop systems with scalable backplanes. The 10xN specification, currently under development, will be backward compatible with the RapidIO Gen1 and Gen2 systems deployed in the market today. Initially, it will support over 10 Gbaud per serial lane with lane widths up to x16, thus allowing for data rates up to 160 Gbps per port. The specification will also scale to serial lane speeds of 25 Gbps and above, as there are no limitations in the logical or transport layer, thus providing continuity as mainstream PHY technology advances. A key aspect for supporting applications is a short-reach specification for local interconnect up to 20 cm and a long-reach one that will support up to 1m, both over two connectors on FR4. Protocol efficiency is further improved by the use of an industry leading coding scheme that reduces the 25% overhead of 8b/10b encoding to less than 5%. The details will be released when the specification is made public. The new roadmap is a response to OEM needs for scaling beyond 6.25 Gbaud per serial lane. The proliferation of 4G handsets is increasing demand on the wireless network. OEMs need more distributed coverage in smaller form factors with more data passing between processing elements, all with low power and ultra low latency. Meanwhile, cloud computing demands a scalable, energy efficient fabric within the data center. Creating scalable systems with 100 Gbps connections to computing nodes, along with ultra low latency to enable high volume financial transactions, is key. High volume financial systems gain a competitive edge from having the lowest and most deterministic latency. Traditional defense and aerospace customers continue to rely on RapidIO for mission-critical systems and increased automation of unmanned vehicles. The 10xN specification, with a path to 25xN, sets a long term roadmap to satisfy the real-world needs of all these applications. The 10xN standard offers major advantages over other protocols. RapidIO does not suffer the software overhead, high latency, and high processor power consumption that occur when scaling other protocols to 10 Gbps per lane and beyond. The new technology roadmap will allow RapidIO to keep supporting its existing customer base, as well as competing in new markets such as servers and cloud computing. Full release of the 10xN specification is planned for late 2011. It is being developed with broad support from all key RapidIO semiconductor, tool, and system vendors and builds on widespread traction in the wireless, defense, aerospace, imaging, video, and cloud computing markets. Tom Cox is Executive Director of the RapidIO Trade Association. You can reach him at tom.cox@rapidio.org. AdvancedTCA Newsletter
5. Feature ArticlesFemtocells Enable New Mobile Connectivity Options The continual demand for higher rates over wireless networks is fueling HSPA+ and LTE technology rollouts worldwide. Femtocells, low-power wireless access points, are one way to provide access in places not suited for traditional wireless deployments and to offload congested macro cells. They use residential DSL or cable broadband connections and can change the mobile broadband industry’s approach to cell design and deployment. Femtocells operate in the licensed spectrum and offer many benefits to both consumers and operators, including increased upload/download performance, better coverage (5 bars), and traffic offload from congested macro cells. ABI Research forecasts a global femtocell market of over 45 million units within five years. Several Tier 1 European mobile operators, including Vodafone and Telefonica, are already actively pursuing femtocell deployments. Furthermore, major players such as Vodafone and T-Mobile, have announced upcoming B2B offerings. According to Informa Telecoms and Media’s 2010 Femtocell Market Status Report, as of December 2010 there were 18 commercial services and a total of 30 deployment commitments. Deployments have almost tripled in the past year. As with all emerging technologies, femtocells must overcome serious challenges before products and services become fully viable. Major issues include scalability, security, mobility, ease of deployment, and interoperability. Proper test solutions in these areas will prevent product and service delays that could impede market growth. Assessing all network aspects and femtocells both individually and as part of an end-to-end system gives equipment manufacturers and operators the ability to meet all deployment challenges. All nodes in the femtocell ecosystem, including user equipment, home eNodeBs (HeNBs), HeNB gateways (HeNB GW), and HeNB management systems, must be thoroughly tested to ensure the highest possible quality. LTE-based femtocell validation must:
According the Yankee Group, 65 percent of carriers will implement some LTE coverage using femtocells “from the bottom up” starting in late 2012. Informa expects the femtocell market to reach just under 49 million Femtocell Access Points (FAP) by 2014, with 114 million mobile users accessing networks through femtocells. Such figures represent a significant investment in this rapidly emerging technology. Joseph Zeto is Senior Manager, Market Development at Ixia. You can reach him at jzeto@ixia.com. Intelligent Traffic Management Solves Video Transport Problems Network operators must provide adequate infrastructure for the huge recent increase in video consumption and the resulting data traffic. Advances in smartphone screen resolution and the emergence of tablets have accelerated the demand for high quality mobile video. In fact, users take HD-level quality as the norm. Offloading is a popular way to deal with massive amounts of data traffic. However, when video is involved, the network must be optimized for distribution, and the process starts in the core. It requires implementing intelligent traffic management by steering and shaping via deep packet inspection (DPI). However, this evolved packet core solution must also scale, which is why AdvancedTCA is the best platform on which to deploy it. Whether the demand is for 10 Gbps or 200 Gbps, network equipment providers (NEPs) must deliver solutions that provide enough bandwidth to ensure operators receive optimal performance, even at peak usage times with bandwidth-hungry HD video traffic. Today’s AdvancedTCA platforms can deliver up to four times the processing density of previous generations’ offerings. The AdvancedTCA form factor enables NEPs to quickly and cost-effectively produce the scalable DPI applications operators need to monitor and control network traffic while ensuring quality of service and experience (QoS/QoE). AdvancedTCA also enables more ubiquitous video optimization platforms that use transcoding and transrating to improve QoE dramatically. With the right combination of packet processing blades, an AdvancedTCA chassis can become a complete 200G DPI platform, enabling operators to scale up their traffic management without having to rip out and replace their entire core network infrastructure. All they have to do is insert new or additional blades. According to a recent report from Multimedia Intelligence, the worldwide mobile video market will exceed $15 billion by next year. However, mobile network operators are currently handcuffed by having to operate within spectral restrictions and a defined radio interface. They cannot rely on fiber to solve the problem. Managing astronomical video traffic growth is a critical requirement for operators, and intelligent traffic management is a route they should consider. To keep users happy, wireless operators must accept alternatives to simply trying to offload traffic from the macro network. Optimizing video via DPI based on AdvancedTCA platforms is a compelling place to start. Jason Byrne is Senior Product Line Manager at Continuous Computing. You can reach him at jbyrne@ccpu.com. Providing Telephone Service During LTE Migration Three themes define today’s 4G / LTE landscape. First, service providers are building mobile broadband access to the internet. Second, networked economies are emerging. And last, new smartphones are really personal computers, capable of handling both data and voice. These themes require high-speed, broadband mobile data networks. 4G/LTE network buildout must address the dramatic rise in mobile data traffic. Studies presented at Mobile World Congress 2011 show that subscribers will pay for a better, richer, faster experience. Higher bandwidth and more content are worth the money. If we build it, they will use it and pay for it. However, voice still provides most current Network Service Provider (NSP) revenues, and LTE requires not-so-ubiquitous VoIP. How will NSPs meet the exploding demand for high-quality, video/data-oriented bandwidth while still providing basic voice, roaming, and text services? Hold off fielding LTE and forego new revenue streams? Not likely. Here is where Circuit Switched Fallback (CSFB) comes into play. It is an idea whose time has come. Back when everyone was touting the ‘All-IP Network’, some observers saw the difficulties inherent in achieving this vision. They are not insurmountable, but they do complicate the required major network overhaul. And the biggest of them is handling voice. CSFB resolves LTE’s lack of backward compatibility with circuit-switched services. All-IP LTE rollouts must include VoIP, or they are simply not practical. At first, this issue was supposed to be addressed by the VOLGA (Voice Over LTE/GAN) initiative, which was deemed too narrow and proprietary. Then came VoLTE (Voice over LTE). Unfortunately, it will take time to offer the same roaming and connectivity as 3G. Meanwhile, CSFB allows network operators to carry voice traffic over existing GERAN/UTRAN networks from multi-mode LTE UE devices. This practical goal is realized by a clever innovation: network awareness in the MME. Where overlapping networks exist, the MME may carry maps of TAs to LAs that allow the UE to utilize circuit-switched services all managed from the MME. If no VoIP services are available, the UE will access the alternate network for voice calls. The CSFB standard states explicitly that there need be no change to User Plane transport services. However, it would be even better if we could combine CSFB with the IP/EP Core. The solution is User Plane interworking at the Network Edge. This approach brings CS-based voice calls into existing VoIP networks quickly and efficiently. In the UTRAN, interworking between ATM-IP is performed on already AMR-encoded voice. For truly traditional voice over DS0s, the AdvancedTCA I-TDM specification allows T1/E1 and DSP cards to pass traffic seamlessly between them and their associated networks. Adax, Surf Solutions, and RadiSys have recently delivered such solutions for both GERAN and UTRAN networks. Broadband mobile access to the Internet is well underway. The future for wireless networking never looked brighter, even in these dark economic times. It’s great to be part of an industry that forges ahead with world-changing applications, devices, services, and networks while still providing the practical solutions needed to maintain current revenue streams. Drew Sproul is Director of Marketing at Adax. You can reach him at dsproul@adax.com. Developing High-Performance Memory Systems Explosive growth of Internet connected devices is currently driving demand for ever faster transmissions and higher bandwidth. Processor speeds have increased according to Moore’s Law, but the performance of memory devices has lagged. At GHz speeds, the number of memory devices needed for buffering rises so high that industry pundits use the term “sea of memory.” Requiring up to 14 chips on some boards, the buffering creates a design bottleneck due to the high pin counts required for the parallel interface. The sea of memory, by definition, also causes a problem because of increased power demand, often stretching the power budget to the verge of bankruptcy at 100G. Still another issue is the memory bank’s cycle time (tRC). At 100G, the network transfers 150 million packets per second (Mpps). As part of its routine operation, network processors update flow management tables monitoring traffic statistics every time a packet arrives. At 150 Mpps, packets arrive every 6.67ns, and the processor requires two memory accesses in this period, or a tRC of 3.33 ns. Meanwhile, every 3.33 ns, or a tRC of 1.65 nsConventional DDR-3 memory, by comparison, specifies tRC at 45 ns. Industry analysts note that, even though 100G developments are progressing, almost all support 10 x 10G. This means that all new 10G boards must still process the 100G data stream. Thus, limited I/O due to current parallel interfaces, memory architectures, and slow tRCs poses an interface and “sea of memory” problem even for new 10G boards. Any solution requires simultaneous advances in interface and memory architecture technologies. For an interface solution, serial is the only viable long-term option. Older serial transmission protocols provide for bus-to-bus transmissions in large packets. Implemented for device-to-device transmissions, the protocols cause highly inefficient transfers, which impose a cost including high power consumption. However, GigaChip, a new high-speed freely available serial interface, transmits packets efficiently to a serial-configured memory array with a low tRC. (See the protocol comparison table below.) It resides on top of the CEI-11 physical standard.
GigaChip syndicates multiple parallel outputs into 72 bit device-to-device transfers. Furthermore, at 100G, a typical DDR-3 memory configuration requires over 1,700 pins, whereas the serial solution requires just 264 total pins (including all power and grounds). Power consumption drops radically as well. To learn more, see http://www.gigachipalliance.com. To summarize, efficient high-speed memory systems require a matching serial interface and memory architecture. GigaChip, when coupled with a memory partner such as the MoSys Bandwidth Engine®, achieves 4x the memory performance of today’s devices and 2x versus other devices sampling this year. Efficient memory performance at 100G requires low pin counts and high throughput, while still staying within the power budget. Serial interfaces optimized for efficient small packet device-to-device transfers and optimized memory answer today’s call for high performance memory subsystems. Michael Sporer is Director of IC Marketing at MoSys. You can reach him at msporer@mosys.com. 6. UNH-IOL Interoperability News and EventsThe University of New Hampshire InterOperability Laboratory (UNH-IOL) reports the following event: October 17, 2011 – October 21, 2011: Ethernet Alliance EEE Plugfest - Ethernet Alliance The Ethernet Alliance Ethernet in the Data Center Subcommittee plans to host another IEEE 802.1 Data Center Bridging (DCB) interoperability plugfest event at the University of New Hampshire Interoperability Lab (UNH-IOL). For more information, visit www.iol.unh.edu. Our thanks to Communications Coordinator Jason Walls for providing this update. He can be reached by phone at (603) 862-5051 or by email at jwalls@iol.unh.edu. 7. Financial NewsEricsson Acquires Telecordia Radisys Closes Acquisition of Continuous Computing Radisys Second Quarter For fiscal year 2012, Total Next Generation Communication Networks revenues are expected to increase by approximately 40% from the prior year and are expected to become at least two-thirds of total Company revenues. Commercial revenues are expected to be down by approximately 25% and Legacy Communications Networks revenue is expected to continue to decline and be down up to 50% from the prior year. In total, revenues are projected to be at a similar level to revenues in 2011. Kontron Second Quarter Freescale Second Quarter NEI Third Quarter 8. Newsletter Advertising Rates
9. Newsletter ArchiveBack issues of this Newsletter are archived in PDF format. See: www.atcanewsletter.com, then click on Archived Newsletters, then click on desired issue.
|
|||||||||||||||||||||||||||||||||
