Archive for August, 2017

Supermicro Delivers Groundbreaking 18 Million IOPS of Storage Performance in New 2U Ultra Server

Thursday, August 24th, 2017

New 1U and 2U All-Flash Ultra SuperServers support 20 Hot-Swap NVMe SSDs with non-blocking Gen 3 PCI-E x4 direct connections for Maximum Performance and Supermicro density-optimized 1U JBOF supporting 32 hot-swap NVMe

SAN JOSE, Calif., Aug. 23, 2017 /PRNewswire/ — Super Micro Computer, Inc. (NASDAQ: SMCI), a global leader in compute, storage, and networking technology and green computing, today announced the availability of new all-flash NVMe™ (Non-Volatile Memory Express) 1U and 2U Ultra SuperServers. With support for 20 hot-swap NVMe SSDs, Supermicro’s new all-flash 2U Ultra server delivers a groundbreaking 18 million IOPS of storage performance.

Supermicro launches 1U and 2U all-flash 20 NVMe Ultra servers and new 32 NVMe 1U JBOF

Supermicro launches 1U and 2U all-flash 20 NVMe Ultra servers and new 32 NVMe 1U JBOF

“To achieve the lowest possible latency, Supermicro’s new all-flash 1U and 2U Ultra servers are designed to support 20 directly attached hot-swap NVMe SSDs,” said Charles Liang, President and CEO of Supermicro. “These new X11 servers feature a non-blocking design, allocating 80 PCI-E lanes to the 20 NVMe SSDs for uncompromised Gen 3 PCI-E x4 direct connections that achieve maximum storage performance.”

With double the number of directly attached hot-swap NVMe SSDs compared to the previous generation, Supermicro’s new all-flash 1U Ultra server supports 20 hot-swap NVMe SSDs. In addition, the performance-optimized new all-flash 2U Ultra unleashes up to 18 million IOPS of storage performance.

Supermicro X11 Ultra servers fully support the highest end Intel® Xeon® Scalable processors up to 205 watts and 24 DIMMs making these servers an excellent choice for high-performance analytics and in-memory application acceleration. The system architecture is balanced to make optimal use of system resources with each processor supporting 10 NVMe drives and dual 25G ports or a 100G port.

For customers looking for even greater increases in the NVMe storage capacity per rack unit, Supermicro’s new 32 hot-swap NVMe SSDs in 1U JBOF (just a bunch of flash) provides maximum high-performance storage density.

With over 100 NVMe based platforms in its X11 server and storage portfolio, Supermicro continues to extend its technology innovation leadership position in NVMe server and storage solutions.

For more information on Supermicro’s all-flash NVMe server solutions, please visit https://www.supermicro.com/products/nfo/NVMe.cfm.

For complete information on Supermicro® SuperServer® solutions, visit www.supermicro.com.

Follow Supermicro on Facebook and Twitter to receive their latest news and announcements.

Contact Information

Super Micro Computer, Inc

980 Rock Ave.
San Jose, CA, 95131
USA

tele: 408-503-8000
Marketing@Supermicro.com
http://www.supermicro.com/index.cfm

Microchip Extends Custom Programming Service to Include AVR® and SAM Microcontrollers

Monday, August 21st, 2017

Manufacturer-Direct Programming is Now Available for More Than 30 AVR and SAM MCU
Families Through microchipDIRECT

CHANDLER, Ariz., Aug. 21, 2017 — Microchip Technology Inc. (NASDAQ: MCHP), has expanded its custom programming service to include AVR® and SAM microcontrollers (MCUs). For the first time ever, users can add their custom code to MCUs from more than 30 AVR and SAM families, along with nearly all PIC® MCUs and memory devices, directly from the manufacturer via microchipDIRECT. Microchip is the only company in the industry that provides an online custom programming service to all of its clients. For more information about this custom programming service, visit: www.microchipdirect.com/avr-sam-programming.html

microchipdirect-avr-and-sam-mcu-custom-programming

Microchip’s custom programming service, which has been widely successful with PIC microcontrollers and memory devices over the past decade, is available to any client regardless of their order size and can be used throughout the development process. From very small runs to verify that the code is working, all the way up through full-scale production runs, this cost-effective programming service offers customers the flexibility to add their code to any order size, from one device to millions. Additionally, each first verification order is complimentary and includes three free samples programmed to each client’s exact specifications.

To get started, clients simply choose their part number on microchipDIRECT and then add their code and other configuration settings, shown on the intuitive online form, directly into the encrypted website. The MCUs will then be programmed directly by Microchip with no need to involve a third party programming or manufacturing facility, thus eliminating the risk of code exposure during the programming process.

“No other manufacturer makes it easier or faster for clients to securely add custom code to their microcontrollers and memory devices,” said Mitch Little, vice president of worldwide sales and applications at Microchip. “Being able to both program and purchase devices through microchipDIRECT truly makes it a one-stop shop that greatly enhances the purchasing experience for our clients while decreasing their risk of code exposure.”

In addition to custom programming services, microchipDIRECT also offers value-added services such as tape and reeling, labels, ink dotting and more. With the largest inventory of Microchip products in the industry, microchipDIRECT provides a full service channel for all purchasing needs. The mobile-optimized website also offers global support in ten languages, volume pricing, live service agents, numerous payment options and order notifications for customer convenience. For more information or to begin purchasing, visit www.microchipdirect.com.

Custom programming directly from Microchip is available for nearly all PIC MCUs and more than 30 AVR and SAM families with additional device support rolling out over the next year. For a full list of supported devices and pricing, please visit: www.microchipdirect.com/programming/CPNPricingFrame.aspx

About Microchip Technology
Microchip Technology Inc. (NASDAQ: MCHP) is a leading provider of microcontroller, mixed-signal, analog and Flash-IP solutions, providing low-risk product development, lower total system cost and faster time to market for thousands of diverse customer applications worldwide. Headquartered in Chandler, Arizona, Microchip offers outstanding technical support along with dependable delivery and quality. For more information, visit the Microchip website at www.microchip.com

Contact Information

Microchip Technology Inc.

2355 W. Chandler Blvd.
Chandler, AZ, 85224
USA

tele: 480.792.7200
toll-free: 888.MCU.MCHP
fax: 480.792.7277
here2help@microchip.com
www.microchip.com

Marvell’s Industry-Leading 88SS1074 SATA SSD Controller Surpasses 50 Million Shipment Milestone in Just 18 Months

Wednesday, August 9th, 2017

Strong customer demand across client, enterprise and cloud data center applications outpacing market

Marvell (NASDAQ:MRVL), a leader in storage, networking, and connectivity semiconductor solutions, continues to accelerate its leading position in the merchant market for solid-state drive (SSD) controller devices. Marvell is proud to announce that its 88SS1074 SATA SSD controller has surpassed 50 million unit total shipments in just 18 months, representing year to date, year over year growth of 385% and outpacing market growth for this category. This significant milestone underscores Marvell’s close working relationships with leading SSD manufacturers, server and storage original equipment makers (OEMs) and cloud data center operators to help develop these solutions with leading performance, endurance and reliability for retail, client, enterprise and data center applications.

Solid state data storage is forecasted to continue its rapid growth and Marvell is well positioned to enable this trend. According to the analyst firm, Gartner, the Mainstream PC-Grade SSD Units market is expected to see a compound annual growth rate of more than 20% from 2016-20211. Building on its legacy of more than 20 years’ experience in storage hard disk drive (HDD) controller technology, complex system-level system-on-chip (SoC) design, advanced error correction algorithms, and low-power architectures, Marvell brings a unique set of capabilities and technologies to address the increasing demands required to enable future SSD solutions with next generations of 3D and QLC NAND technology. In addition, Marvell has recently developed a set of robust, full turnkey (FTK) SSD controller solutions for multiple new generations of NAND components and SSD features to help accelerate the time to market for customers.

“We are excited to continue our long-standing collaboration with Marvell to provide industry-leading client and data center SSD products,” said Ariel Perez, SSD business manager, Kingston. “We’ve seen first-hand the advanced error correction capability that Marvell’s NANDEdge™ controller technology provides to enhance NAND flash endurance and reliability, making it ideal for our next-generation 3D NAND drive series.”

“High performance, high endurance and low power consumption are all critical SSD requirements for the client, enterprise and cloud data center markets we serve,” said Darlo Perez, managing director, Americas Region, Storage Strategic Business Group at Lite-On. “Marvell’s SSD controllers are best-in-class and we’re excited to continue our long-time collaboration with Marvell to satisfy the burgeoning demand for highly efficient SSDs.”

“Micron’s customers have the most demanding enterprise and cloud data center workloads and require SSD technologies that lower the total cost while delivering high performance and reliability, superior data protection, and optimal endurance,” said Eric Endebrock, vice president SSD and Systems, Storage Business Unit, Micron Technology, Inc. “We congratulate Marvell on their contributions to innovative SSD solutions and on reaching this important milestone.”

Marvell has been investing in SSD controllers since 2007, and released its first SSD controller in 2009. Marvell is now on its fifth generation of SATA-based controllers and its fourth generation of PCIe®-based ones. Marvell’s decades of research and development, coupled with its culture of innovation, has resulted in a robust IP portfolio of over 1400 storage patents granted and approximately 200 pending.

“Marvell is extremely proud that in collaboration with our customers, our 88SS1074 6G SATA SSD controller has surpassed 50 million unit shipments,” stated Dan Christman, executive vice president, Storage Business Group at Marvell. “We continue to aggressively invest in advancing our industry-leading technology to meet the demands of the accelerating client-to-cloud data storage SSD market. Our growing portfolio of solutions and partnerships positions us at the forefront of this opportunity, and we look forward to attaining many more milestones with our customers in the future.”

The total data storage capacity enabled by 50 million controllers is staggering. For context, the number of photographs taken by the global population each year is approximately 1.3-1.5 trillion. Assuming that pace continues, 50 million SSDs using Marvell’s SATA controllers could easily store all photographs taken by the entire global population in the coming five years.

A key factor in securing these record-breaking shipments has been the introduction of Marvell’s proprietary NANDEdge technology. This low-density parity check (LDPC) error-correction mechanism, which is unmatched by anything else currently available in the industry, is incorporated into all of Marvell’s latest SSD controllers spanning SATA, SAS and NVMe™. Suitable for both conventional planar and emerging 3D stacked triple level-cell (TLC) and quad level-cell (QLC) NAND SSDs, NANDEdge technology ensures the highest degrees of endurance, reliability and data integrity are maintained and prolongs SSD operational lifespan.

Marvell is showcasing its next-generation architecture solutions for client, industrial, automotive, cloud and enterprise applications, including its latest SSD controllers, at the Flash Memory Summit 2017 in Santa Clara, Calif. from August 8-10. Scott Furey, associate vice president of Marvell’s Enterprise Storage Business Unit, will present a conference keynote today from 2:40-3:10 p.m. on “Using Arm®-Based Processing for Efficient Hyperscale Storage.”

For further information on Marvell’s storage solutions, please visit: http://www.marvell.com/storage

1 Gartner, Forecast: Hard-Disk Drives, Worldwide, 2014-2021, 1Q17 Update, 15 May 2017

About Marvell
Marvell first revolutionized the digital storage industry by moving information at speeds never thought possible. Today, that same breakthrough innovation remains at the heart of the company’s storage, networking and connectivity solutions. With leading intellectual property and deep system-level knowledge, Marvell’s semiconductor solutions continue to transform the enterprise, cloud, automotive, industrial, and consumer markets. To learn more, visit: www.marvell.com .

Contact Information

Xilinx Showcases Reconfigurable Storage Acceleration Solutions at Flash Memory Summit 2017

Tuesday, August 8th, 2017

Xilinx, Inc. (NASDAQ: XLNX) today announced it will showcase reconfigurable storage acceleration solutions at the Flash Memory Summit 2017. Through a series of presentations and demonstrations, Xilinx and its ecosystem will spotlight high-performance storage solutions for use in current and next-generation enterprise and data center applications. Debuting at this year’s event, the Xilinx® NVMe-over-Fabrics reference design provides designers a flexible platform to enable scalable storage solutions and integrate custom acceleration functions into their storage arrays. This reference design eliminates the need for a dedicated x86 processor or an external NIC, thus creating a highly integrated, reliable and cost-effective solution. To learn about this solution and more, visit Xilinx at booth #126, August 8-10, 2017 at the Santa Clara Convention Center, Santa Clara, California.
Presentations/Panel Participation:
Thursday, August 10:

10:15-10:30AM: Reconfigurable Storage Acceleration for Next Generation NVMe Platforms by Rakesh Cheerla, Product Manager, Xilinx
1:30-4:15PM: Using FPGAs to Accelerate NVMe-over-Fabric Based Storage Networks by Deboleena Minz Sakalley, Senior Staff Design Engineer, Xilinx
1:30-2:45PM: Accelerating Data Analytics Using FPGAs and an Integrated Flash Storage by HK Verma, Principal Engineer, Xilinx
Xilinx Booth Demonstrations

NVMe-over-Fabric Platform
The Xilinx single-chip storage solution integrates NVMe-over-Fabric and target RDMA offloads with a processing subsystem to provide a very power-efficient and low-latency solution compared to existing products that require both an external host chip and a Network Interface Card (NIC). This 2x100Gb Ethernet platform enables customers to implement value-added storage workload acceleration, such as compression and erasure code.
Computational Storage Subsystem for Data-Driven Applications
ScaleFlux Computational Storage Subsystem (CSS) uniquely solves both compute and storage I/O bottlenecks. CSS compression accelerates throughput by an order-of-magnitude without the costly CPU overhead incurred by software solutions, resulting in optimal storage capacity utilization without performance compromises.
Programmable Controller for Multi-Source Flash
Burlywood’s TrueFlash modular controller architecture accelerates time-to-market for new NAND adoption while delivering disruptive cost and performance benefits to cloud, all-flash-array, and hyper-converged OEM customers. Built on Xilinx® UltraScale+™ FPGAs, TrueFlash takes advantage of the power, performance and cost improvements of Xilinx devices. TrueFlash enables customers to quickly optimize the solution to their workload, and to qualify multiple NAND types with the same controller.
Demonstrations by Xilinx Ecosystem
The following Xilinx ecosystem companies will showcase a Xilinx technology-based demonstration in their FMS booth:

Everspin Technologies: Booth #319
IntelliProp, Inc.: Booth #821
IP-Maker: Booth #717
Kazan Networks: Booth #702
MobiVeil: Booth #610
PLDA: Booth #826
Smart IOPs: Booth #609
About Xilinx
Xilinx is the leading provider of All Programmable semiconductor products, including FPGAs, SoCs, MPSoCs, RFSoCs, and 3D ICs. Xilinx uniquely enables applications that are both software defined and hardware optimized – powering industry advancements in Cloud Computing, 5G Wireless, Embedded Vision, and Industrial IoT. For more information, visit www.xilinx.com.

Contact Information

Xilinx

2100 Logic Drive
San Jose, CA, 95124
USA

tele: 408-559-7778
more_info@xilinx.com
www.xilinx.com

UCloud Selects ScaleFlux for Data-Driven Cloud Infrastructure

Monday, August 7th, 2017

Computational Storage Deployment Signals a Sea Change in IT Infrastructure Solutions to Address the Crashing Wave of Compute and Storage I/O-Intensive Applications

SAN JOSE, Calif.–(BUSINESS WIRE)–ScaleFlux, Inc., the pioneer in deploying Computational Storage at scale, announced that UCloud, China’s leading independent public cloud provider, has selected ScaleFlux to power its advanced UPHost (UCloud Physical Host) infrastructure for Big Data, Database, Data Warehousing, and a fast-growing list of data-driven applications.

“One of the primary values of CSS is the ability to reprogram new hardware acceleration engines in-system, which can evolve the infrastructure in lock step with new customer demands.”
Tweet this
ScaleFlux provides industry’s first, production-qualified Computational Storage Subsystem (CSS) solutions that uniquely integrate algorithmically complex compute engines with Terabytes of 3D NAND flash storage. CSS simply integrates into standard server and storage infrastructure through an easy-to-install Linux plug-in software package combined with standard PCIe HHHL Add-In Card and 2.5” U.2 drive level form factors.

“UPHost offers outstanding computing performance to meet the core application scenarios for high performance and stability, and is built to run applications at any scale for our customers,” said Xianfeng Mo, CTO and Founder of UCloud. “With the integration of ScaleFlux CSS into UPHost, we are further extending UCloud’s value proposition by dramatically reducing the time to complete Big Data and Database jobs, while also optimizing flash storage capacity utilization.”

Large pools of Computational Storage will eliminate the scaling roadblocks faced by data-driven infrastructure where:

(1) traditional CPUs are highly inefficient when tasked to process modern, intense workloads

(2) the latency and power to move massive datasets from storage to the CPU is too expensive

(3) compute-intensive workloads can be sped up by orders-of-magnitude when parallelized across flash storage subsystems

These scaling challenges are commonly faced with applications in the field of Big Data, Database, Data Warehousing, Content Delivery, Search, Machine Learning, AI, VR, Genomics, Security, High-Performance Computing, and many others.

“By taking an evolutionary and simple approach to integrating CSS into application frameworks, we are delighted to provide our revolutionary Computational Storage solutions for UCloud’s UPHost infrastructure that will open new revenue generating use-cases for data-driven applications at scale with fast time-to-market,” said Hao Zhong, CEO and co-founder for ScaleFlux, Inc. “One of the primary values of CSS is the ability to reprogram new hardware acceleration engines in-system, which can evolve the infrastructure in lock step with new customer demands.”

ScaleFlux CSS features field-programmable gate arrays (FPGAs) and enables acceleration for compression, erasure coding, and KV store, in addition to encryption, authentication, and common compute intensive algorithms. Future application level acceleration algorithms such as database queries, machine learning, AI, media transcoding, search, and many more, can be dropped into the same hardware infrastructure. By taking a programmable hardware approach, ScaleFlux has set CSS apart from commodity SSDs through the addition of proven, application level, value-added features while facilitating mass deployment of state-of-the-art 3D NAND Flash media.

ScaleFlux CSS is available for customer deployment today. More information is located at www.scaleflux.com.

About ScaleFlux, Inc. (www.scaleflux.com)

ScaleFlux is the pioneer in providing Computational Storage Subsystem (CSS) solutions at scale. CSS is the foundation for modern, data-driven infrastructure that provides responsive performance, affordable scaling, and agile platforms for compute and storage I/O intensive applications. Founded in 2014, ScaleFlux is a well-funded startup with leaders proven to deploy complex computing and solid-state storage solutions in volume.

About UCloud (www.ucloud.cn)

UCloud (Shanghai UCloud Cloud Computing Co., Ltd.) is a leading IaaS service provider in China. UCloud’s own R&D solutions range from computing and storage to network resources and other fundamental IT infrastructure that enterprises need. UCloud has provided services to 50,000 enterprise customers worldwide.

Contact Information

ScaleFlux

97 East Brokaw Road
Suite 260
San Jose, CA, 95112
USA

info@scaleflux.com
http://www.scaleflux.com/

Smart IOPS Announces Two New Products To Complement Its Impressive High-Performance NVMe Solid-State Drive Portfolio

Monday, August 7th, 2017

Data Engine Low Latency Edition accelerates caching, enables more latency-sensitive and high transaction workloads per server. The Data Engine High Capacity Edition increases the storage density of hyper-converged systems, hyperscale nodes, storage and All-Flash Arrays.

SANTA CLARA, Calif., Aug. 7, 2017 /PRNewswire/ — Smart IOPS, Inc. a leading flash storage solution provider announced today that it is expanding its already impressive portfolio of high-performance Data Engine PCI Express (PCIe) NVM Express (NVMe) enterprise solid state drives (SSDs). In all, the company is set to offer three new and distinctive categories of data center and High–Performance Computing (HPC) SSD products targeting multiple market segments:

Data Engine T2HP High-Performance: The fastest and highest performing PCIe NVMe SSD available in the market today. Offered in 3.2 and 6.4TB user capacities, Data Engine T2HP is in a class of its own, surpassing 1.7 million IOPS and 6.8GB/s in bandwidth per SSD.
Data Engine T2HC High Capacity: One of the largest user capacity in a half height half length (HHHL) add-in card form factor with up to 25.6TB. Unlike other larger SSDs, the performance in this product continues to far exceed any other SSD in its general class.
Data Engine T2LL Low Latency: Smart IOPS has combined the superior performance that it has come to be known for with best-in-class NAND technology to rival the latest non-volatile memory products offered by others. However, unlike those products that require forklift server upgrade, Data Engine T2LL offers QoS latency and response time that is on par and at a small fraction of the cost of those new trendy non-volatile memory solutions. Moreover, Data Engine T2LL offers random and sequential Read/Write performance that is orders of magnitude greater than those caching non-volatile memory products.
“Our patent-pending truRandom™ technology offers such a diversified set of services that when combined with different 3D NAND flash technologies, it can address a multitude of market sub-segments and tackle storage challenges at a more granular level,” said Ashutosh Das, CEO of Smart IOPS. “Our goal is very simple – we intend to offer the highest IOPS, greatest throughput and lowest latency at budget-friendly prices. We do not believe data centers should have to upgrade their entire server infrastructure with the latest processor and chipset in order to get outstanding performance. You might say that Smart IOPS in the business of connecting dots, not crossing points.”

Smart IOPS is sampling Data Engine T2 HP to a select group of OEMs and enterprises. Data Engine High Capacity and Low Latency editions are expected to sample in CQ4’2017. For additional information about Smart IOPS Data Engine SSDs, please visit smartiops.com.

ABOUT SMART IOPS

Smart IOPS is an innovative leader in storage solutions whose truRandom™ technology creates new possibilities for flash technology to have a profound impact on current and next generation data center applications. Powered by truRandom, Smart IOPS SSDs substantially increase IOPS and bandwidth to remove the I/O storage bottleneck that severely limits the modern data center. Smart IOPS’ flash storage solutions enable Cloud Service Providers (CSP), High-Performance Computing (HPC) and enterprise data centers to confidently deploy I/O-intensive applications with greater reliability and SLA, at a fraction of the cost of other flash-based storage products. Visit the company website at www.smartiops.com for more information.

Contact Information

Smart IOPS

1551 McCarthy Boulevard
Suite 208
Milpitas, CA, 95035
USA

tele: 408-216-8100
fax: 408-216-8100
http://www.smartiops.com/

Faster Storage and Networking Interfaces Boost Performance of COM Express and CompactPCI Serial-based Systems

Thursday, August 3rd, 2017

Opening new applications for Transportation, Machine Control and Test.

The adoption of new storage form-factors, such as DOM, M.2 and U.2, which employ the wider use of non-volatile memory express (NVMe) architecture, are fueling new industrial applications. NVMe technology improves performance when building small systems that require large, fast local data storage.

Second, as system-on-module (SOM) solutions mature, next generation Ethernet interfaces are being offered as integrated Network Interface Controller (NIC) chips included within the SOM device. More full-featured computer-on-module (COM) solutions that use 10 Gb capable integral NIC devices are coming on line.

As companies including congatec and ADLINK roll out the COM Express Type 7 module (Figure 1), more capable small form-factor systems can now be built. New systems can now offer better, 10 Gbit Ethernet network interfaces as part of the COM Express module.

Figure 1: Typical COM Express Type 7 Module

Figure 1: Typical COM Express Type 7 Module

More Technological Advancements

Considering other system solutions that require pluggable or removable boards, advancement in follow-on Eurocard pluggable modules is demonstrated with new chips being offered on CompactPCI Serial modules, instead of the older parallel bus-based CompactPCI standard.

These standards are controlled by the PICMG organization, which makes them readily accessible to users for a small fee. As more suppliers come into the CompactPCI Serial space, better chip level solutions allow more functions to be placed on a 3U Eurocard.

A key differentiator with CompactPCI Serial, over the older CompactPCI, is that the new form-factor supports a high-speed module-to-backplane differential pair connector, allowing the user access to a standardized pluggable 3U module that can mate with a mid- to high-speed backplane.

Mechanical and Interface Design vs. Function

Considering the storage, network, and packaging needs, a quick look at form-factor, size, and electrical interfaces reveals:

  • Changes to the physical storage form-factor
  • Introduction of new physical network interfaces and connectors
  • Fanless designs being used with COM Express modules

Further, 3U CompactPCI Serial plug-in modules or boards employ serial fabric-based interfaces that now support low slot count, high performance, and lower-cost 3U board-based systems.

Figure 2: Storage Modules: M-SATA vs. M.2

Figure 2: Storage Modules: M-SATA vs. M.2

Storage

These devices are changing from the 2.5” and 1.8” solid state disk (SSD) form-factor to new choices developed for the laptop market. Over the last five years, Mini-PCIe and M-SATA have been introduced and used as new form-factors on COM Express carriers and various motherboards (Figure 2).

These edge card-based modules have been placed on COM Express carriers and within motherboards to expand I/O and provide storage solutions in a smaller space. More recently, M.2, and now U.2, SSDs have emerged that also use edge card connectors, but are offered in larger foot prints than Mini-PCIe or M-SATA modules.

These modules are offered with solid state storage capacities ranging from 128 MB to 2 TB. More interesting is the shift to the use of the NVMe storage interface. NVMe uses the PCIe electrical standard, resulting in much higher storage bandwidth.

Considering a one lane interface comprised of a send and receive pair, the older SATA 3 interface would allow 1.5, 3.0, or 6 Gb/sec transfers. NVMe with PCIe allows 2.5, 5, and 8 Gbaud transfers, providing 2.0, 4.0, and effectively just under 8 Gb/sec throughputs per lane.

With NVMe, devices can use one, four, eight, or more lanes that can scale the storage device interface bandwidth—allowing the physical connection to the SSD array to operate at GB/sec throughputs instead of MB/sec throughputs. To achieve higher rates, new SSD memory devices from Samsung, Intel, and Micron now allow SSDs to operate at two to four times the transfer rate of SSDs that formerly used the 6 Gb SATA 3 interface.

A similar trend is seen with NVMe-based 2.5” SSDs that now use the U.2 form-factor and allow similar performance increases.

Network Interfaces

The industry now supports the use of USB 3.0 and 10 Gb Ethernet on Type 6 and Type 7 COM Express modules, respectively. The Type 7 COM Express modules features two 10 Gb Ethernet interfaces. The use of these newer high-speed interfaces requires better board design and improved connectors to accommodate the higher data rates.

Figure 3:  Storage Module Carrier with 4 TB on CompactPCI Serial 3U Board

Figure 3: Storage Module Carrier with 4 TB on CompactPCI Serial 3U Board

Specifically, considering the most recent change to the COM Express specification, the module definition supports both 1 and 10 Gigabit Ethernet interfaces. Small servers can now be built using COM Express modules. Further, the new USB 3.0 and Ethernet interfaces enable applications that utilize high-resolution video formats, such as HD-SDI and H.265.

Broadwell DE class CPUs provide integral 10 Gb Ethernet interfaces, along with traditional 1 Gb network interfaces. Considering the speed, manufacturers must consider whether to offer copper-only interfaces or to add the option to provide fiber interfaces.

In either case, presenting high speed I/O to the outside world requires new connectors.

OnTime Networks is offering embedded routers that provide I/O through Ceelok Fas-T connectors to distribute 10GBASE-T Copper Interfaces. Likewise, Elma Electronic Inc. has moved to the Hercules Connector to bring out higher-speed signals in newer COM Express based small form-factor systems.

High-speed Connection

Looking at other advances from PICMG draws attention to adoption of the CompactPCI Serial form-factor. First introduced in 2011, the form-factor was more quickly adopted in Europe.

As new CPU and network devices are offered, there is a greater need to place these chips in front of higher-speed connectors to support the board-to-board connectivity. CompactPCI Serial does just that, using higher speed differential connectors on the module and the backplane. CompactPCI Serial allows industrial users of the older parallel-based CompactPCI standard to migrate to a serial fabric-based module and backplane scheme suited to industrial use at lower cost points that fit this market segment.

Typical applications include rail and transportation computer systems as well as machine control and now test instruments. For many, the draw to CompactPCI Serial has increased due to the use of standard Eurocard chassis with off-the-shelf backplanes, providing a pluggable and expandable modular system approach.

Manufacturers, such as Kontron, EKF, and MEN Micro, offer a wider variety of CompactPCI Serial boards to fit cost-sensitive, moderately rugged applications that can make use of the 3U Eurocard format (Figure 3). Standards-based packaging is available from existing manufacturers, with a choice of suppliers for the chassis and backplane solutions.

This can make implementation more straightforward with CompactPCI Serial vs. a COM Express solution, since standard packaging exists for CompactPCI Serial modules.

Compact Systems Enable High Data Throughput

New applications such as revised communication gateways and hyper-spectral image processing systems, as well as video analysis and presentation systems, can be realized by hosting them on small form-factors, or lower-cost, standard modular systems, such as CompactPCI Serial.

Network interfaces providing 10 Gb network connectivity allow the higher capacity data links necessary to support the transmission and processing of high-definition video, and handle aggregate data rates found in multi-sensor based systems.

New NVMe-based storage solutions align with higher-resolution sensors, and higher-performance CPUs, to support data capacity and the data rates emerging applications require.

The bits and pieces needed to implement smaller, faster embedded computing platforms continue to evolve, allowing developers to bring better platforms to market that host and enable exciting new embedded applications.


KenGrobKen Grob is Director of Embedded Technology, Elma Electronic. He holds a BSEE from Drexel University and has been in the embedded computing industry since 1985. The company manufactures electronic chassis platforms, embedded boards for storage, networking and processing, provides system architecture design, development, and implementation of complex solutions for communications, military/aerospace, medical and semiconductor equipment manufacturers.

Accelerates AI Evolution: AccelStor Introduces Next Generation High Availability All-NVMe Flash Array

Wednesday, August 2nd, 2017

AccelStor NeoSapphire H810 Featuring Uncompromising Performance at Flash Memory Summit 2017

FREMONT, Calif.–(BUSINESS WIRE)–AccelStor, the software-defined all-flash array provider, announced a new high availability all-NVMe flash storage array, the NeoSapphire H810, accelerating the AI era with its uncompromising performance, endurance and capacity topped with AccelStor’s FlexiRemap® flash-oriented software technology. The NeoSapphire H810 will help customers to remove their system bottlenecks by offering a truly seamless user experience. AccelStor’s all-flash array products and technology will be on display at the Flash Memory Summit 2017, Booth #132, in Hall A of the Santa Clara Convention Center, in Santa Clara, California, USA, from August 8 to 10, 2017.

[News_photo]_AccelStor's_Next_Generation_High_Availability_All-NVMe_Flash_Array
“The digital transformation is happening at an incredibly rapid pace with the ascent of the connected device, big data analysis and artificial intelligent evolution. These trends are creating tremendous opportunities and also challenges for us to deliver seamless, simple, sophisticated solutions,” said Dr. Weafon Tsao, AccelStor Vice President, “We are extremely excited about the new edge of NVMe flash. Integrating AccelStor’s FlexiRemap® technology with NVMe platform means we are able to live up to the potential of our storage solution for our performance-hungry enterprise customers with extreme low latency and maximized throughput, and furthermore, to assist them explore new business opportunities.”

Dr. Weafon Tsao, will be giving a presentation at the Flash Memory Summit, presenting “NVMeoF Enterprise All-Flash Arrays Offer Top Performance” on August 9th sharing AccelStor’s vision and ambition about the latest NVMeoF flash technology.

Ultra-high Performance and Low Latency

By leveraging the latest generation of the Intel Xeon Scalable Processor, the Purley platform increases up to 1.6x workload-optimized performance versus the prior generation. The NeoSapphire H810 is an all-NVMe flash array as 4U rack mount delivering over 6 million IOPS for 4KB random access. It delivers an overall performance increase of up to 600% compared to the previous NeoSapphire P310 all-NVMe flash array, accelerating today’s modern-day workloads including artificial intelligence, deep learning, virtualization, HPC applications and digital content creation. The H810 comes with 100GbE connectivity supporting the demand from network admins for the highest data rates over a single connection.

True High Availability of NeoSapphire H810

Thanks for the FlexiRemap® technology and shared-nothing architecture, the H810 accomplishes true high availability with no single point of failure. Its symmetric active-active failover design and real-time data synchronization offers microsecond-level access latency and the essential zero downtime.

This approach, is specifically designed to provide no single point of failure and minimal service interruptions, and we believe it is the way to true high availability. To synchronize the data between 2 nodes, AccelStor adopts a special approach for data transfer. AccelStor takes the conventional RDMA (Remote Direct Memory Access) technology to the next level. To transfer the data from node A to node B, AccelStor’s patented technology will pack random 4KB I/O as a capsule. During the capsule transfer, it will efficiently reduce I/O latency and enable fast message synchronization with minimal CPU overhead.

Even under a massive random I/O workload, AccelStor’s unique high-availability extension software, combined with the backbone of the low-latency InfiniBand interlink, enables highly efficient data transfer between the two nodes. Therefore, incoming data can be synchronized in real time and safely preserved on both nodes.

Total Package for Data Service Software Features

Powered by AccelStor’s FlexiRemap® software technology, the NeoSapphire H810 delivers abundant features for data service, advanced redirect-on-write snapshot provides both data recovery and an efficient method for capacity and performance management. The H810 also packs AccelStor’s Free Clone feature, which enables copying of datastores in the vSphere environment without allocating physical capacity. Other handy tools include snapshot backup/recovery, remote replication, inline compression, FlexiDedupe, and the vSphere Web Client plug-in. Even with these rich software features, the H810 still boosts performance for computationally intensive tasks, making data more accessible for advanced analytics.

Additional Resources

Learn more about the High-Availability NeoSapphire series
Learn more about FlexiRemap® Technology
Follow AccelStor on Twitter: @AccelStor
Follow AccelStor on Linkedin

About AccelStor, Inc.

AccelStor is accelerating the paradigm shift from conventional disk arrays to modern all-flash storage. AccelStor’s NeoSapphire all-flash arrays, powered by FlexiRemap® software technology, deliver sustained high IOPS for business-critical applications. With streamlined storage management, multi-protocol support, and front-access, hot-swappable solid-state drives, the NeoSapphire series promises to resolve performance bottlenecks for I/O-intensive applications like virtualization, high-performance computing, database, and media processing. For more information about AccelStor and NeoSapphire, please visit www.accelstor.com.

Contact Information

AccelStor, Inc.


New Taipei City ,
Taiwan

www.accelstor.com

Engineers’ Guide to Embedded Storage

Tuesday, August 1st, 2017

cover_98