HYPER-CONVERGED INFRASTRUCTURE

Inspur Announces MLPerf v2.0 Results for AI Servers

Inspur | July 04, 2022 | Read time : 3 min

Inspur
The open engineering consortium MLCommons released the latest MLPerf Training v2.0 results, with Inspur AI servers leading in closed division single-node performance.

MLPerf is the world’s most influential benchmark for AI performance. It is managed by MLCommons, with members from more than 50 global leading AI companies and top academic institutions, including Inspur Information, Google, Facebook, NVIDIA, Intel, Harvard University, Stanford University, and the University of California, Berkeley. MLPerf AI Training benchmarks are held twice a year to track improvements in computing performance and provide authoritative data guidance for users.

The latest MLPerf Training v2.0 attracted 21 global manufacturers and research institutions, including Inspur Information, Google, NVIDIA, Baidu, Intel-Habana, and Graphcore. There were 264 submissions, a 50% increase over the previous round. The eight AI benchmarks cover current mainstream usage AI scenarios, including image classification with ResNet, medical image segmentation with 3D U-Net, light-weight object detection with RetinaNet, heavy-weight object detection with Mask R-CNN, speech recognition with RNN-T, natural language processing with BERT, recommendation with DLRM, and reinforcement learning with MiniGo.

Among the closed division benchmarks for single-node systems, Inspur Information with its high-end AI servers was the top performer in natural language processing with BERT, recommendation with DLRM, and speech recognition with RNN-T. It won the most titles among single-node system submitters. For mainstream high-end AI servers equipped with eight NVIDIA A100 Tensor Core GPUs, Inspur Information AI servers were top ranked in five tasks (BERT, DLRM, RNN-T, ResNet and Mask R-CNN).

Continuing to lead in AI computing performance

Inspur AI servers continue to achieve AI performance breakthroughs through comprehensive software and hardware optimization. Compared to the MLPerf v0.5 results in 2018, Inspur AI servers showed significant performance improvements of up to 789% for typical 8-GPU server models.

The leading performance of Inspur AI servers in MLPerf is a result of its outstanding design innovation and full-stack optimization capabilities for AI. Focusing on the bottleneck of intensive I/O transmission in AI training, the PCIe retimer-free design of Inspur AI servers allows for high-speed interconnection between CPUs and GPUs for reduced communication delays. For high-load, multi-GPU collaborative task scheduling, data transmission between NUMA nodes and GPUs is optimized to ensure that data I/O in training tasks is at the highest performance state. In terms of heat dissipation, Inspur Information takes the lead in deploying eight 500W high-end NVIDIA Tensor Core A100 GPUs in a 4U space, and supports air cooling and liquid cooling. Meanwhile, Inspur AI servers continue to optimize pre-training data processing performance, and adopt combined optimization strategies such as hyperparameter and NCCL parameter, as well as the many enhancements provided by the NVIDIA AI software stack, to maximize AI model training performance.

Greatly improving Transformer training performance

Pre-trained massive models based on the Transformer neural network architecture have led to the development of a new generation of AI algorithms. The BERT model in the MLPerf benchmarks is based on the Transformer architecture. Transformer’s concise and stackable architecture makes the training of massive models with huge parameters possible. This has led to a huge improvement in large model algorithms, but necessitates higher requirements for processing performance, communication interconnection, I/O performance, parallel extensions, topology and heat dissipation for AI systems.

In the BERT benchmark, Inspur AI servers further improved BERT training performance by using methods including optimizing data preprocessing, improving dense parameter communication between NVIDIA GPUs and automatic optimization of hyperparameters, etc. Inspur Information AI servers can complete BERT model training of approximately 330 million parameters in just 15.869 minutes using 2,850,176 pieces of data from the Wikipedia data set, a performance improvement of 309% compared to the top performance of 49.01 minutes in Training v0.7. To this point, Inspur AI servers have won the MLPerf Training BERT benchmark for the third consecutive time.

Inspur Information’s two AI servers with top scores in MLPerf Training v2.0 are NF5488A5 and NF5688M6. The NF5488A5 is one of the first servers in the world to support eight NVIDIA A100 Tensor Core GPUs with NVIDIA NVLink technology and two AMD Milan CPUs in a 4U space. It supports both liquid cooling and air cooling. It has won a total of 40 MLPerf titles. NF5688M6 is a scalable AI server designed for large-scale data center optimization. It supports eight NVIDIA A100 Tensor Core GPUs and two Intel Ice Lake CPUs, up to 13 PCIe Gen4 IO, and has won a total of 25 MLPerf titles.

About Inspur Information
Inspur Information is a leading provider of data center infrastructure, cloud computing, and AI solutions. It is the world’s 2nd largest server manufacturer. Through engineering and innovation, Inspur Information delivers cutting-edge computing hardware design and extensive product offerings to address important technology sectors such as open computing, cloud data center, AI, and deep learning. Performance-optimized and purpose-built, our world-class solutions empower customers to tackle specific workloads and real-world challenges.

Spotlight

Alstom, the world leader in transportation solutions, modernizes its infrastructure to cut maintenance costs for IT resources to 10 percent of the original budget, reduce the time to provision new environments from months to minutes, and improve cooperation of teams across the globe.


Other News
IT SYSTEMS MANAGEMENT

Stream and T-Systems Partner to Empower Advanced Hybrid Cloud Architecture

Stream Data Centers | August 02, 2022

Stream Data Centers, the industry leader in delivering exceptional data center experiences to global enterprise companies, is proud to announce that it has been selected as a trusted data center partner by T-Systems. T-Systems, a leading provider of Information and Communication Technologies (ICT) solutions to major corporations and public-sector organizations across the globe, chose Stream Data Centers' Houston campus in the Woodlands to host customers and help deliver innovative cloud operation services both locally and globally. As part of Deutsche Telekom Group (DT), T-Systems is a leading digital and cloud services provider, offering world-class service while supporting local and global customers by extending its global portfolio, expertise, and operational capabilities. Led by its dedication to providing transformative ICT solutions, the company today has nearly 100 managed data centers, 56,800 open system servers, and more than three million managed SAP users. While expanding its newest hybrid cloud platform, T-Systems found that it needed a strategically-located data center that could cater to its architecture's high-density demands as well as the company's own ESG (Environmental, Social and Governance) goals. "Our mission is to provide the best solutions to our customers, with the right partners, using state-of-the-art technology," states Mauro Guzelotto, Head of Cloud Services for T-Systems. "It was in the spirit of that mission that we decided to partner with Stream, and we are very excited about the value and possibilities that Stream brings to the table. Another key element of our decision was sustainability, and I am confident that our decision to partner with Stream will contribute to our sustainability strategy by helping us be more energy efficient." After a rigorous RFP process, Stream and its Houston campus were selected for a host of reasons, including the location's advantageous geographical setting, Stream's ability to meet sustainability goals with its energy procurement and operational expertise, and additional strategic service offerings including high-density capabilities. The Houston I facility in the Woodlands is located outside of the 500-year floodplain and has 185 mph wind ratings with an uplift-rated building and equipment yard. These aspects have enabled this facility to offer 100% power and cooling uptime for the last 8 consecutive years — even standing strong against 1,000-year storms like Hurricane Harvey and Winter Storm Uri. It is also connected to a separate power grid from T-Systems' core facility, which allows for added redundancy. As a secondary site, Stream's campus offered superior benefits and assurance against downtime. Stream also enabled T-Systems to benefit from a partnership with Megaport, enriching its public cloud connections and further enabling the delivery of a robust hybrid cloud platform. T-Systems' tailored cooling and power-per-rack demands could also be easily met with Stream's large rooms, 3-foot raised floors and tall ceilings, which help the ambient temperature remain easily controlled even for high-density deployments. Furthermore, T-Systems' multi-stage goals for sustainability across its operations (with an ultimate goal of fully eliminating its carbon footprint) are empowered by this facility's LEED certification and the Stream team's insights into strategic energy procurement and efficient usage. "With this symbiotic partnership in place, Houston-area customers can enjoy T-Systems' leading suite of IT service offerings and leverage their innovative platforms to further their own digital transformation initiatives. "Being selected by a global leader like T-Systems is a great testament to Stream's Woodlands facility and team members." Chad Rodriguez, Vice President of Network and Cloud at Stream Data Centers "The partnership between T-Systems North America and Stream Data Centers allows our joint existing and future customers to travel their cloud journeys with the certainty that availability, reliability, sustainability, and security are managed by two of the most relevant IT and DC experts in the market," comments Cesar Martinez, Managing Director for T-Systems in North America. About Stream Data Centers Stream Data Centers has provided premium data center services since 1999, with 90% of its inventory leased to Fortune 100 customers. To date, the company has acquired, developed and managed 24 data center campuses nationally, while leadership has remained consistent for all 23 years.

Read More

DATA STORAGE

SANBlaze Enters New Markets in the Storage Testing Industry

SANBlaze | August 01, 2022

SANBlaze Technology Inc., a leading worldwide provider of advanced storage test and validation technologies, today announced the expansion of its industry-first NVMe® over PCIe® 5.0 validation and compliance testing system from traditional SSD manufacturers to new markets comprised of data center storage and large cloud vendors. “Customer confidence has grown beyond our traditional walls of satisfying requirements from major SSD manufacturers to supporting large data centers and cloud storage organizations. “This evolution stems from our first-to-market leadership for early adoption and development of NVMe PCIe Gen5. Early availability was a critical factor in enabling our key strategic customers to meet their internal development schedules for Gen 5 SSD’s and FCS releases.” Rick Walsh, VP of Sales and Marketing “SANBlaze partnered with WD to get our Gen5 Validation infrastructure ready in time including SRIS/SRNS clocking features which helped to fast track our overall Gen5 bring up,” said Anuj Awasthi, Senior Director, System Design and Firmware Verification Engineering, Western Digital Corp. In addition to SSD manufacturers such as Western Digital, SANBlaze is onboarding new major cloud and data center storage providers as they recognize the capabilities and value of the Certified by SANBlaze test suites as a first-level SSD validation criterion. Certified by SANBlaze is an instant benchmarking tool that saves on CapEx overhead expenses for SSD compute applications. SANBlaze Suite of Products SANBlaze solutions include the SBExpress-RM5™ rackmount appliance, the SBExpress-DT5™ desktop appliance, and the industry benchmark SBExpress Certified by SANBlaze software, which provides over 900 ready-to-go tests and scripts. These latest PCIe 5.0 platforms provide broad test capabilities for development, QA, validation, and manufacturing teams in data centers large and small. SBExpress-RM5 The SBExpress-RM5 is a 16-bay enterprise-class NVMe test appliance supporting hot-plug and all link speeds up through PCIe 5.0. The system features a unique modular “riser” design that enables user-configurable variable slot support, as well as field-upgradable support for all PCIe 5.0 connector form factors, including U.2, M.2, EDSFF, and the new E3/EDSFF. The ability to margin and measure power, glitch signals, and test spread spectrum clocking (SSC) or conventional clocking in both common and SRIS/SRNS modes sets the SBExpress-RM5 apart from all others in the NVMe SSD testing space. Data integrity is verified with a comprehensive suite of read/write/compare tests, with exception cases such as power glitching while running IO, and built-in "Write Atomicity" testing as part of the Certified by SANBlaze test suite. Testing is accessible through a web interface to the appliance or via Python, XML and REST APIs, which come standard with the system. The SBExpress™ Gen5 software includes over nine-hundred test scripts to enable IOL testing in the customer’s lab, before undergoing official testing, as well as ZNS, VDM, and TCG Opal verification. SBExpress-DT5 The SBExpress-DT5 is the sixth-generation SBExpress system and is both evolutionary, growing from its successful family of predecessors, and revolutionary, with advanced test capabilities such as Vendor Defined Messaging (VDM) testing, MI (Management Interface) in-band, and SMBUS testing at 1MHz. All features of the enterprise test suite Certified by SANBlaze are supported by DT5 at PCIe 5.0 speed. SANBlaze at Flash Memory Summit Flash Memory Summit 2022 takes place August 2-4 at the Santa Clara Convention Center, Santa Clara CA, USA. SANBlaze, a member of the Symbiosys Alliance, will be present in booth #219. The Symbiosys Alliance will be present in booth #119. About SANBlaze SANBlaze is a pioneer in storage testing and validation technologies. SANBlaze systems are deployed in the test and development labs of most major storage hardware and software vendors worldwide. SANBlaze is revolutionizing the NVMe Storage Area Network (SAN) and PCIe device qualification markets by offering NVMe testing end-to-end. We are first to market a solution that tests Native NVMe and NVMe over Fabrics (NVMe-oF™) for complete end-to-end testing of your entire system using single port or dual port drives. About the Symbiosys Alliance The Symbiosys Alliance is an I/O interconnect technology group chartered to create value for its membership and for their respective customers by strategically and collaboratively aligning member products and services to current and upcoming market opportunities. These synergized solutions can provide developers with the state-of-the-art resources they need to roll out highly competitive offerings efficiently and confidently to their respective marketplaces. The alliance addresses a range of verticals increasingly characterized by hyper-fast innovation cycles. These include semiconductors, data storage, IoT, cloud computing, consumer electronics, automotive, aerospace, medical, and more. Members leverage alliance partnerships to precisely anticipate and address these innovation cycles by delivering high-quality solutions that resonate with the latest technological advances.

Read More

IT SYSTEMS MANAGEMENT

Keysight Enables Celona to Accelerate 5G Private Network Deployments

Keysight Technologies | May 20, 2022

Keysight Technologies, Inc. , a leading technology company that delivers advanced design and validation solutions to help accelerate innovation to connect and secure the world, announced that Celona has selected Keysight’s Open RAN Architect (KORA) solution portfolio to validate the quality and reliability of 5G private network deployments for enterprises. Keysight’s integrated test, validation and optimization tools enable Celona to accelerate the deployment of 5G local area network (LAN) solutions, including access points for indoor and outdoor wireless coverage and cloud-native software powered by artificial intelligence (AI). As a result, Celona can deliver solutions and services to the private mobile network sector with software and hardware components that speed private wireless network deployments and maximize available network capacity. “Keysight is pleased to support Celona in delivering end-to-end networking solutions that seamlessly integrate with existing enterprise networks to simplify private cellular wireless operations and accelerate the adoption of business-critical applications requiring deterministic wireless connectivity. “Keysight’s test solutions speed industrial device and network validation and certification, network planning, deployment and site acceptance, as well as simplify network monitoring critical in smart manufacturing, energy, utilities, mining, transport and health care.” Giampaolo Tardioli, vice president and general manager for Keysight's enterprise and service providers group Celona serves a growing private network market with turnkey 5G LAN solutions that are purpose-built for enterprises. Celona delivers deterministic and reliable connectivity, security and wireless performance by integrating cellular wireless communications within existing enterprise-owned IT infrastructure. “Our enterprise customers expect high quality and performance of deployed critical wireless infrastructure,” said Mehmet Yavuz, co-founder and chief technology officer at Celona. “Keysight’s test tools enable Celona to quickly verify the 5G radio operation and performance of Celona’s 5G LAN solutions, leading to significant reduction in time to market and improved performance.” Celona is using several Keysight 5G radio access network (RAN) test tools, including: Keysight’s user equipment emulation (UEE) test solution (UeSIM) to emulate real network traffic over radio and O-RAN fronthaul interfaces, enabling comprehensive base station performance validation across the protocol stack. Keysight's S9130A 5G Performance Multi-Band Vector Transceiver (VXT), a non-signaling measurement system, to accelerate the validation of sub-6GHz and mmWave 5G base stations (gNodeBs) according to the latest 3GPP specifications. About Keysight Technologies Keysight delivers advanced design and validation solutions that help accelerate innovation to connect and secure the world. Keysight’s dedication to speed and precision extends to software-driven insights and analytics that bring tomorrow’s technology products to market faster across the development lifecycle, in design simulation, prototype validation, automated software testing, manufacturing analysis, and network performance optimization and visibility in enterprise, service provider and cloud environments. Our customers span the worldwide communications and industrial ecosystems, aerospace and defense, automotive, energy, semiconductor and general electronics markets. Keysight generated revenues of $4.9B in fiscal year 2021.

Read More

IT SYSTEMS MANAGEMENT

CoreSite Expands into Atlanta and Orlando Data Center Markets

CoreSite | June 07, 2022

CoreSite, a leading hybrid IT solutions provider and subsidiary of American Tower Corporation , today announced its expansion into the Atlanta and Orlando markets with the integration of three American Tower assets into CoreSite’s data center ecosystem: Atlanta AT1 (55 Marietta Street NW, Atlanta, GA, 30303), Atlanta AT2 (1130 Powers Ferry Place, Marietta, GA, 30067) and Orlando OR1 (9701 S. John Young Parkway, Orlando, FL, 32819). All three data centers, which have a combined total of more than 250,000 square feet, now also offer access to the Open Cloud Exchange® (OCX), CoreSite’s leading software-defined networking platform which provides fully managed, direct and secure connections into all of the major cloud service providers. The Atlanta and Orlando data centers, former DataSite and Colo-ATL assets acquired by American Tower, provide businesses of all sizes the secure, reliable interconnection solutions to meet their critical infrastructure requirements. The OCX provides superior connectivity and a highly interconnected partner network needed to reach new markets, rapidly scale on-demand, reduce total cost of operation and accelerate IT modernization. “With a long-standing track record of operational excellence in these markets, these facilities have been serving Atlanta- and Orlando-area businesses for years and now operate as an integral part of the broader data center ecosystem within CoreSite. “The Atlanta and Orlando teams are equipped with the technical, security, remote hands and customer service expertise organizations rely on to future-proof their digital businesses. Offering low-latency, redundant and secure colocation services as well as the OCX, businesses of all sizes can seamlessly and dynamically scale their IT infrastructure to meet their ever-changing needs.” Juan Font, President of CoreSite and SVP, U.S. Tower Division of American Tower As one of the fastest-growing data center markets, Atlanta has the third-largest concentration of Fortune 500 companies in the United States. CoreSite’s Atlanta data center campus includes AT1, which is located in the second most interconnected building in the heart of downtown Atlanta, and AT2 in Marietta. Together these assets are strategically positioned to serve this rapidly growing technology hub. Orlando, a city whose goal is to be a “Future-Ready City,” offers enterprises, IT services providers and cloud service providers (CSPs) a densely populated market that is positioned to be a digital transformation leader. CoreSite’s Orlando campus is a 16-acre safe haven with regional and global reach that enables access via internet exchange to South America. The integration of these assets into the CoreSite data center portfolio expands upon CoreSite’s offering to now include 27 data centers in 10 markets, 450+ networks, 23 native cloud onramps and 35,000+ interconnections. “This Atlanta and Orlando market expansion delivers on the value proposition originally envisioned with the CoreSite acquisition,” said Steve Vondran, Executive Vice President and President, U.S. Tower Division of American Tower. “With the incorporation of these three assets, CoreSite and American Tower continue to develop the infrastructure required to offer seamless, end-to-end connectivity between mobile data networks at the tower and digital platforms at the data center campus. The complementary global communications real estate portfolio and financial wherewithal that American Tower brings will continue to accelerate our growth and leadership in the emerging 5G digital ecosystem.” About CoreSite CoreSite, an American Tower company (NYSE: AMT), provides hybrid IT solutions that empower enterprises, cloud, network, and IT service providers to monetize and future-proof their digital business. Our highly interconnected data center campuses offer a native digital supply chain featuring direct cloud onramps to enable our customers to build customized hybrid IT infrastructure and accelerate digital transformation. For more than 20 years, CoreSite’s team of technical experts have partnered with customers to optimize operations, elevate customer experience, dynamically scale, and leverage data to gain competitive edge.

Read More

Spotlight

Alstom, the world leader in transportation solutions, modernizes its infrastructure to cut maintenance costs for IT resources to 10 percent of the original budget, reduce the time to provision new environments from months to minutes, and improve cooperation of teams across the globe.

Resources