Huawei Reveals a Highly Efficient Big Data Appliance Prototype, NÜWA, at Spark Summit 2015
Jun 17, 2015
[San Francisco, USA, June 17, 2015] Huawei has unveiled a prototype of its all-in-one big data appliance named NÜWA at the 2015 Spark Summit in San Francisco. The NÜWA appliance integrates computing and storage and has a unique design which focuses on the co-optimization of hardware and software components. Current test results show that NÜWA improves price-performance ratio over 10 times when compared with the open-source Spark running on standard computer servers, providing a truly innovative big data solution for customers with higher performance requirements.
Currently, many enterprises assemble their big data platforms from existing components. With soaring data volumes and increased business lines, pressure on storage capacity and computing ability is mounting. Power consumption and space limitations, make it difficult for larger enterprises to upgrade and expand their data centers. To meet these challenges, Huawei has made multiple innovations for the architectural design of big data appliances, and has developed the all-in-one NÜWA appliance which integrates computing, storage and analytics.
NÜWA is the industry's first micro-server with disaggregated architecture completely based on commercial chips and the computing and storage nodes can be switched on demand. While NÜWA helps simplify management, it also achieves over 50% power saving when compared to ordinary computer servers, and the appliance is over 75% smaller than ordinary in-class servers. Equipped with the industry's highest-performance data engine with dynamic compilation execution technology, NÜWA implements an automatic query switch that shifts from execution through ordinary function calls to execution through on-the-fly compilation. This competes with fully optimized User Defined Function (UDF) codes. According to test results, the data processing speed of NÜWA is between 10 and a 100 times faster than the open-source Spark. NÜWA adopts a special Erasure Coding (EC) algorithm for faster speeds, lower recovery costs, and storage savings of over 50% when compared to the Hadoop distributed file system (HDFS). In addition, with the unique Field Programmable Gate Array (FPGA) âDataflow Processorâ and the data analysis execution plans that are generated intelligently, NÜWA optimizes and executes all Structured Query Language (SQL) queries with performance better than the existing FPGA acceleration solutions.
'The Huawei NÜWA appliance is able to cover the typical scenarios of both private and public clouds, with the prerequisites of high-speed processing, compatibility and usability, and on-line expandability for the hardware and software. The NÜWA prototype has been completed and Huawei has commenced customer pilots with renowned companies, and plans to launch the product officially at the end of this year. We hope to provide many customers with the cost-effective NÜWA big data appliance solutions for a win-win result,' said Wei Zhulin, NÜWA project leader, Huawei.
The name of NÜWA comes from a mythological figure in ancient China who repaired the pillar of heaven with stones and made mankind from yellow clay. The name has been selected to represent that Huawei uses ordinary chips to develop a revolutionary product and to redefine the IT server market through integrated and innovated designs and optimized algorithms.