working solutions. So, we're going to talk about um accelerating AI fabrics for storage. So, you know, AI, we we put a lot of effort into accelerating it and improving it uh GPUs, uh you know, from from the entire stack.And one of the things we find is that slow storage can slow down your AI training times. It'll slow down your uh your uh inference. Like even with things like rag uh there's a lot of work that's done to uh deal with the ingest phase where it's like it's kind of reading the material and so th ...

如果您希望可以时常见面，欢迎标星收藏哦~ 来源：内容 编译自 theregister 。 在英特尔将其Omni-Path互连技术剥离给Cornelis Networks五年后，这家公司终于准备好凭借其 400Gbps CN5000系列交换机和网卡（NIC）正面对抗老对手Nvidia的InfiniBand技术。 这一次，Cornelis的目标不仅限于超级计算机和高性能计算（HPC）集群，它还希望借助更高的性 价比切入AI热潮，与Nvidia展开竞争。 对于那些早已将Omni-Path抛诸脑后的读者，以下是快速回顾：Omni-Path最初由英特尔于2015年 开发，是一种无损互连技术，在许多方面与Nvidia的InfiniBand网络相似，主要面向高性能计算应 用。 第一代Omni-Path交换机提供4.8Tbps带宽，共有48个100Gbps端口，曾部署于多个超级计算平 台，如洛斯阿拉莫斯国家实验室的Trinity系统和美国能源部的Cori系统。 但 到 了 2019 年 ， 英 特 尔 放 弃 了 该 项 目 ， 并 于 2020 年 9 月 将 该 业 务 部 门 剥 离 成 立 Cornelis Netw ...

如果您希望可以时常见面，欢迎标星收藏哦~ 来 源：内容 编译自 servethehome 。 Cornelis Networks 在英特尔剥离 Omni-Path 技术后接手了这项技术，如今该公司推出了新一代 400Gbps 产品。Cornelis CN5000 将 400Gbps Omni-Path技术带入人工智能和高性能计算 (HPC) 市场。同时，该公司分享了未来愿景，计划在下一代产品中采用 UltraEthernet 技术。 400Gbps 网络发展迅速。其中一大推动力是人工智能的扩张。更快的 AI 网络意味着大量 GPU 可 以更快地发送和接收任务，从而获得更高的利用率。没有什么比能够提高数十亿美元 GPU 的利用 率更能体现投资回报率了。 新款 Cornelis CN5000 采用 400G SuperNIC 设计（SuperNIC 也是NVIDIA和其他公司使用过的 术语）。它可以连接到 48 端口交换机或最多 576 端口的导向器交换机。Cornelis 为生态系统制 造了 SuperNIC 和交换机 ASIC。 这一代的导向器级交换机不仅采用风冷，还配备了液冷版本。我们听到多家大型AI公司讨 ...

Summary of Conference Call Company and Industry - **Company**: Arista Networks - **Industry**: Networking and Cloud Infrastructure Key Points and Arguments AI Opportunities - Arista is fully committed to AI, transitioning from experimental phases to production with cloud customers [3][4] - The company categorizes its AI initiatives into two areas: Networking for AI (high-speed, low-latency products) and AI for Networking (using AI/ML for observability and security) [4][5] - The AI market is seen as a continuation of existing cloud opportunities, with a total addressable market (TAM) of $70 billion, including $20 billion each for data center, cloud, and enterprise markets [11][12] Market Dynamics - The relationship between AI and cloud is symbiotic, with AI clusters increasing traffic demands on cloud infrastructure [12] - Arista's software stack is a significant differentiator, applicable to both cloud and AI environments, emphasizing open networking principles [14][15] - AI traffic is characterized by higher speeds and different flow patterns compared to traditional cloud traffic, necessitating advanced network management [20][21] Customer Engagement and Competition - Arista has strong relationships with key hyperscale customers like Meta and Microsoft, but faces challenges with Google and Amazon, who prefer white box solutions [21][22] - The company acknowledges the presence of white box competitors but maintains its leadership in spine networking and offers integration with white box solutions [26][27] Technology and Product Differentiation - The Jericho chip family is crucial for managing congestion in AI workloads, with 80% of Arista's deployments utilizing this technology [30][32] - Arista supports pluggable optics and is exploring co-packaged optics, emphasizing reliability and troubleshooting capabilities [36][40] - The company sees a growing need for automation and visibility in network management, which is critical for handling AI traffic [28][38] Financial Guidance and Management Changes - Arista reported strong Q1 performance and a robust Q2 guide, with a focus on maintaining momentum despite macroeconomic uncertainties [50][52] - Recent management changes have raised investor concerns, but the company believes it has reached a stable state with a younger, experienced team [56][58] Enterprise and Campus Opportunities - The enterprise and campus markets are becoming more active post-pandemic, with increased demand for integrated wired and wireless solutions [70][71] - Arista is committed to serving all market segments, ensuring equal focus on enterprise, campus, and cloud opportunities [73][74] Additional Important Content - The company emphasizes the importance of Total Cost of Ownership (TCO) for customers, balancing capital and operational expenditures [27] - Arista's unique culture and management style, which combines engineering excellence with customer focus, is seen as a competitive advantage [60][61]