AWS exclusive details of FPGA fundamentals and market development
At the end of the year at the end of 2016, AWS (Amazon Web Services) announced that it can use Xilinx high-end FPGA devices with the help of cloud transport models, and for the first time to help potential users learn and experience from a developer's perspective rather than extending high-level tools. The acceleration effect of the FPGA. Deepak Singh, director of the AWS container and HPC division, introduced to The Next Platform that there are many similarities in the fastest growing application areas, such as the desire to adopt cloud-based FPGA solutions, as detailed in the recently published A book "FPGA Frontiers: New Applications in Reconfigurable Computing." These involve encryption and security, genomics, financial services, and a range of machine learning related applications. “There are already many FPGA use cases for security, genomics and financial services. We are looking forward to seeing the combination of FPGA and machine learning. It will be a broader field, we need to provide more tools and support. Satisfy its growth." Singh focuses on specialized workload research, including high-performance computing, and has been involved in and witnessed technology changes for more than a decade in the AWS department, from generic cloud services (similar to basic bread and butter) to new computing. And other aspects such as memory-intensive applications and high-performance computing (HPC) GPU acceleration solutions. In addition to focusing on the hardware and software requirements for such applications, Signh said he is also focusing on two other areas of broader development and impact, namely machine learning and related hardware requirements. Machine learning is an important area of ​​development for FPGAs, Signh said, but in terms of scale it only tends to specific professional users, and now is not sure which devices the reconfigurable device might be suitable for, "it might be as a GPU, CPU, or as Customized ASICs, tensors, or just as FPGAs, or a combination of certain types," he said, because of the GPUs, programmable environments, and emerging ecosystems in recent years, users can be more open. To explore new architectures and accelerators, although they are still about the same application scenario, at least some users with strong interests (not just for general purpose functions) and his team will always pay attention to the new architecture, and finally choose AWS. solution. Signh also said that the AWS FPGA cloud service solution will attract some users in the early stage, and will also encourage partners to participate in the FPGA technology development process. General-purpose computing won't disappear, and it's still the main force driving computing. But the Nvidia Tesla GPU that has emerged in the last two or three years and its performance make professional programming devices easier, and the interest in professional processors, developers, workloads, and the size and activity of the company are emerging. To a certain extent, specialized hardware and infrastructure play a very large role, especially for those with specific needs, we think FPGA is the best choice. Singh mentioned that many end users have started to move. If the general workloads used are not completed in the cloud, the local data center may be in the past. Almost no electrical equipment can form a single, dedicated workload. "There may be nothing left in addition to FPGA-based devices, because other related ones are already out." For such a situation, he said that the relationship between hardware and device vendors will eventually shift to Amazon cloud services. There are not many vendors with FPGA core machines, but as we have described in the past, Edico Genome (FPGA-accelerated genomic platform) and Ryft (FPGA-driven large-scale data analysis platform) have become good partners in Amazon's efforts. The F1 server instance meets new professional requirements. Buy your "FPGA Frontier: FPGA Frontiers: A New Application for Reconfigurable Computing 2017", Next Platform, available on Amazon and other e-commerce platforms As Singh mentioned, when AWS adds GPU instances, the GPU ecosystem is very mature, and there is still a long way to go for FPGAs to provide a wider range of access services. These advances will allow current vendors and ISVs (independent software developers) to continue to promote, but for AWS, in addition to providing hardware development kits, it will continue to explore to provide more access compliance and permissions. "We plan to support high-level tools such as SDAcells, including the OpenCL library, which expands the resources of developers. There are already some software developers who understand both FPGAs and how to use core tools. They need a channel - they can Their software is available to more customers, and at the same time, we are researching the availability and accessibility of various resource libraries and programming tools to give users better operational performance," Singh explained. "Our The goal is not to make programming and using FPGAs an obstacle." Singh also mentioned the development of FPGA instance cloud services through AWS internal discussions, where team members dedicated to dedicated computing areas (such as HPC, application acceleration, etc.) have witnessed the growth of FPGAs and the ability to meet the computing power of applications in key market segments. Requirements, such as the early adoption of FPGAs as a system solution aspect (financial services, etc.), the adoption of FPGAs as system acceleration (encryption/security, genomes), and the desire to combine various processing computation engines to satisfy various algorithms and Calculate applications with fast load growth (such as machine learning and deep learning). As we mentioned in the previous article, especially in the last two years, more and more professional computing requirements have prompted many companies to no longer be limited to standard CPUs, but gradually adopt ASICs, GPUs and FPGAs. This greatly enriches the diversity of high-end computing solutions and emerges more standard/non-standard system development architectures. Last year, FPGA manufacturer Altera was acquired by Intel, and public cloud services began to adopt FPGA. Various applications will benefit from FPGA acceleration characteristics in all aspects, which is also a major development year for FPGAs - 2017 Will continue to pay attention to this progress! Laser Instrument Shell Packaging Laser Instrument Shell Packaging,Gtms Enclosure With Gold Plating,Glass To Metal Sealed Enclosure With Gold Plate,Glass To Metal Sealed Enclosure Terminal Shenzhen Capitol Micro-Electronics Co.,LTD , https://www.capitolgtms.com