Extending access to HPC manufacturability feedback software through hardware-accelerated virtualized workstations
Extending access to HPC manufacturability feedback software through hardware-accelerated virtualized workstations
复制标题
通过硬件加速虚拟化工作站扩展对 HPC 可制造性反馈软件的访问
DOI：
10.1109/isfa.2016.7790174
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
T. Kurfess
中科院分区：
文献类型：
--
作者：
Roby Lynn;Didier Contis;M. Hossain;Nuodi Huang;Thomas M. Tucker;T. Kurfess
计算机辅助制造（CAM）对许多制造工厂的日常运营至关重要。CAM软件通过将计算机模型转换为机床指令来辅助零件制造。最常见的是，CAM用于生成G代码，这是数控车床和铣床的标准编程语言。虽然CAM软件极大地简化了将零件模型转换为G代码的过程，但要学会正确使用它仍需要经验。此外，运行CAM软件需要一台强大的工作站，由于其成本和体积，这可能会不方便。许多机构已将虚拟化作为多个独立工作站的替代方案；虚拟化允许多个用户访问由远程服务器托管的桌面环境。这具有将用户与某些软件的操作系统和硬件要求隔离开的优点，并且还允许他们在任何地方运行所需的应用程序。本研究探索了图形处理单元（GPGPU）上虚拟化通用计算这一新兴领域；该技术用于支持使用一种新颖且易于使用的CAM软件包，该软件包利用了GPU的并行计算能力。虚拟化可以降低大量用户的实施成本，并降低支持需求，因为GPU和其他计算硬件集中在一个位置。初步结果显示，由于虚拟化中固有的硬件抽象，存在性能损失。然而，与向用户轻松部署CAM系统相比，这些损失就显得微不足道了。
Computer-aided manufacturing (CAM) is essential to the everyday operations of many manufacturing facilities. CAM software is used to aid in the manufacturing of a part by converting a computer model into instructions for a machine tool. Most frequently, CAM is used to generate G-Code, which is the standard programming language of CNC turning and milling machines. While CAM software greatly simplifies the process of converting a part model to G-Code, it still requires experience to learn to use properly. Additionally, a powerful workstation is required to run the CAM software, which can be inconvenient due to its cost and size. Many organizations have been turning to virtualization as an alternative to multiple standalone workstations; virtualization allows for many users to access desktop environments that are hosted from a remote server. This has the benefit of isolating the user from both OS and hardware requirements for certain software, and also allows them to run the applications they need from anywhere. This research explores the emerging area of virtualized general purpose computation on graphics processing units (GPGPU); this technique is used to support the use of a novel and easy-to-use CAM package that leverages the parallel computation capability of a GPU. Virtualization can allow for both lower implementation costs for a large number of users and lowered support requirements, as the GPUs and other computing hardware are consolidated in one location. Preliminary results show performance losses due to hardware abstraction inherent in virtualization. However, these losses are overshadowed by the ease of deployment of the CAM system to users.