So how to integrate with the production environment? How to achieve better application of FPGA advantages in the actual business system? The actual deployment is the most worthy research direction. In this regard, Shenwei technology has made a series of explorations and attempts, and formed the following schemes.

In the cloud, thunderimage is usually integrated with obs. The client sends a request for an image. CDN checks whether there is a local cache. If not, it sends a request to obs. OBS interacts with thunderimage through the corresponding parameters to generate the corresponding image format, and finally returns it to the client through CDN. This way of online processing integrated with OBS brings the reduction of bandwidth cost and storage cost, as well as the improvement of QoS.

At present, FPGA product deployment can be divided into public cloud solution and private cloud solution. At present, the output products of solutions of Shenwei technology to the industry are mainly divided into public cloud: SaaS, instance and image service; Private cloud and offline products: all in one machine. The application scenarios include mobile phone, e-commerce, social networking, CDN, cloud storage, video, etc.

For private cloud deployment, the architecture is based on the hardware environment of video acceleration card, the corresponding operating system on the upper layer of the server and virtualization processing, covering the parts of Xilinx driver layer, management layer tools, etc. The solution of Shenwei technology is deployed on this architecture, directly facing the corresponding application scenarios.

Alveo   The U50 adopts the ultrascale + Architecture and integrates the ultra-high bandwidth 8GB hbm2 memory technology, so the product size is greatly reduced. The solution based on this is the highest computing density solution, which is more convenient for data center server upgrade.

FPGA + CPU image processing solution in data center

The above figure shows an actual case of a social network cloud photo album thumbnail production. The throughput performance of a single node can be improved by 16 times, the TCO can be reduced to half, and it has a better quality of service experience. Because of its smaller cluster size, it is more convenient to maintain.

Public cloud deployment mainly involves platforms, functions and services. The service forms on cloud include SaaS, instance and image. The advantages of SaaS are easy to deploy, low starting cost and convenient for small-scale users to integrate. The integration of instance is more flexible, the encapsulation is more complete, and the cost is lower than SaaS. For image, the purchase of virtual machine and image requires the perfect integration of application and image on the server side, which means that we can have the opportunity to give full play to the best performance of the overall scheme. Therefore, the performance of image mode is the strongest and can achieve in-depth customization.

FPGA + CPU image processing solution in data center

The above figure is an actual case of a video website’s webp transcoding scheme. Our overall performance in the cloud is guaranteed. Although the CPU in the cloud is vcpu, there will be some loss in performance, but the overall performance gain will generally be improved by more than 20 times.

To sum up, the thunderimage image image processing solution of deep dimension technology supports rich encoding and decoding algorithms, can adapt to differentiated application scenarios, and is seamlessly compatible with common algorithm platforms in the industry. Efficient and convenient application experience and deployment mode is a new optimization and choice for public cloud, private cloud and hybrid cloud applications.


Leave a Reply

Your email address will not be published. Required fields are marked *