Photonic Switched Optically Connected Memory: An Approach to Address Memory Challenges in Deep Learning
Photonic Switched Optically Connected Memory: An Approach to Address Memory Challenges in Deep Learning
复制标题
DOI：
10.1109/jlt.2020.2975976
复制
发表时间：
2020-05
影响因子：
4.7
通讯作者：
Ziyi Zhu;G. D. Guglielmo;Q. Cheng;M. Glick;Jihye Kwon;H. Guan;L. Carloni;K. Bergman
中科院分区：
文献类型：
--
作者：
Ziyi Zhu;G. D. Guglielmo;Q. Cheng;M. Glick;Jihye Kwon;H. Guan;L. Carloni;K. Bergman
深度学习已经彻底改变了我们社会的许多方面，为包括计算机视觉、自然语言处理和活动识别在内的各个领域提供了动力。然而，数据集和模型大小的缩放趋势限制了系统性能。内存需求的可变性可能导致资源利用率低下。可重新配置的光子互连提供了可扩展的解决方案，并能够有效地使用分散的存储器资源。我们提出了一种光子交换光连接存储器系统架构，该架构可以解决存储器挑战，同时展示深度学习模型的光交换功能。我们提出的系统架构采用了“建兴”（去）串行化方案的存储器传输通过光链路，以避免网络开销，并支持远程存储器的动态分配到本地处理系统。为了测试我们的建议的可行性，我们建立了一个实验测试床的处理系统和两个远程存储节点使用硅光子交换结构和评估系统的性能。光交换时间为119 μs，端到端重配置的总延迟为2.78 ms。集体的结果和现有的高带宽的光学I/O显示集成的光子交换光连接的存储器的国家的最先进的处理系统的潜力。
Deep learning has been revolutionizing many aspects of our society, powering various fields including computer vision, natural language processing, and activity recognition. However, the scaling trends for both datasets and model size are constraining system performance. Variability of memory requirements can lead to poor resource utilization. Reconfigurable photonic interconnects provide scalable solutions and enable efficient use of disaggregated memory resources. We propose a photonic switched optically connected memory system architecture that tackles the memory challenges while showing the functionality of optical switching for deep learning models. Our proposed system architecture utilizes a “lite” (de)serialization scheme for memory transfers via optical links to avoid network overheads and supports the dynamic allocation of remote memories to local processing systems. In order to test the feasibility of our proposal, we built an experimental testbed with a processing system and two remote memory nodes using silicon photonic switch fabrics and evaluated the system performance. The optical switching time is measured to be 119 μs and an overall 2.78 ms latency is achieved for the end-to-end reconfiguration. The collective results and existing high-bandwidth optical I/Os show the potential of integrating the photonic switched optically connected memory to state-of-the-art processing systems.