Workflow
超融合基础设施
icon
Search documents
博通客户,被抢光了
半导体芯闻· 2025-08-29 10:12
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 CRN 。 "所有这些合作伙伴都选择与我们合作,也带来了部分客户," 他对《CRN》称,"我认为这些合作 伙伴手中的客户数量远多于我们目前已获取的规模。显然,合作伙伴在我们争取这些客户的过程中 发挥了关键影响力。不过,合作伙伴无法在一夜之间将所有客户都转移过来,毕竟客户迁移需要时 间。尽管如此,如今合作伙伴与我们的合作力度已远超以往。" 拉马斯瓦米表示,总部位于美国加利福尼亚州圣何塞的 Nutanix,正处于从 VMware 手中夺取市 场份额的 "第二局"—— 在他看来,这一市场机遇将持续 5 至 10 年。今年新增 2700 家客户后, Nutanix 的客户总量已达 2.9 万家,其中包含 50 多家《全球 2000 强》(指按营收排名的全球大 型企业榜单)企业客户。 该公司预计,在未来整个财年内,新增客户数量将保持中高个位数的增长速度。 "过去一年新增 2700 家客户,这一数据充分表明市场正朝着我们的方向转移," 他在财报电话会 议上对分析师表示,"但 VMware 目前仍有 20 万家客户,因此我们还有大量市场空间需要开拓, 这也需 ...
AI为何成基础设施投资核心驱动力 解读IDC最新报告
Sou Hu Cai Jing· 2025-07-28 09:18
企业级 AI 应用在落地过程中,对 IT 基础架构的性能、资源利用率、容器环境支持、多样化数据存储能 力等方面有较高的要求: l 灵活的计算与存储资源调度:在进行 AI 应用开发时,不同开发组对 GPU 资源的需求量不同,一些开 发任务也不会完全占用一块 GPU 卡的资源;在使用 AI 应用时,不同应用对 GPU 和存储的资源需求也 不尽相同,且需求量可能变化频繁。这些都要求 IT 基础设施能够灵活切分、调度计算与存储资源,同 时支持高性能 CPU 与 GPU 算力,在提升资源利用率的同时满足不同应用/开发任务的资源需求。 l 高性能、低时延的存储支持:对行业大模型进行微调时使用的 GPU 规模较大,要求存储能够为 GPU 并行计算提供高性能、低时延的数据支持。AI 应用的全流程也要面对多个数据源的大量数据读取/写 入:源数据通过预处理,可参与到大模型的微调和推理过程,并对推理形成的文本/语音/视频数据进行 保存和输出。这些工作都要求存储具备高速写入与读取能力。 l 多样化的数据存储支持:上述 AI 应用相关的工作流程中,需要同时处理结构化数据(如数据库)、 半结构化数据(如日志)和非结构化数据(如图像和文本 ...
突破教育科研新格局!摩尔精英联手深信服重磅推出“教学科研一体化平台”,重塑算力想象空间
半导体行业观察· 2025-03-14 00:53
Core Viewpoint - The article emphasizes the launch of a new integrated teaching and research training platform by Moer Elite and Deepin Technology, aimed at addressing the challenges faced by educational and research institutions in the digital transformation era [1][2]. Group 1: Industry Pain Points - The education and research sectors are experiencing significant challenges due to outdated information infrastructure, including hardware silos, insufficient computing power, complex storage systems, and a lack of practical training applications [3][4]. - Hardware silos lead to high operational costs and complexity due to the deployment of servers, storage, and networks from different vendors [3]. - Traditional servers are unable to meet the high computing demands of big data and AI, resulting in inadequate performance for concurrent and multi-tasking needs [3]. - The growth of data has made traditional storage systems difficult to scale, leading to increased management costs [3]. - There is a lack of accompanying research training software, which hampers the ability to provide a comprehensive teaching and practical environment for students and researchers [3]. Group 2: Integrated Platform Features - The new integrated platform combines high-performance computing, elastic storage, and training software to provide a "turnkey" solution for educational and research institutions [3][4]. - Deepin Technology's Hyper-Converged Infrastructure (HCI) consolidates servers, storage, and networks into a unified resource pool, offering lower costs, higher resource utilization, and greater reliability compared to traditional systems [6][7][8][9][10]. - The platform allows for easy deployment and management, enabling educators and researchers to focus on teaching and research rather than hardware maintenance [11]. Group 3: Software Integration - Moer Elite's research training software is designed to work seamlessly with Deepin Technology's hardware, enhancing the platform's capabilities and fostering academic innovation [18]. - The software provides comprehensive training content and resources for various disciplines, including chip design and AI, and supports customized tools for research institutions [19][20]. Group 4: Advantages of the Integrated Machine - The integrated machine offers convenient deployment, flexible expansion, simplified management, high cost-effectiveness, and robust security features [21][22][23][24][25]. - It supports a wide range of educational and research scenarios, from small-scale projects to large-scale simulations, ensuring that institutions can adapt to varying data and computing needs [26][29]. Group 5: Future Outlook - The integrated teaching and research training platform is positioned to significantly reduce deployment and operational costs while enhancing the quality of education and research [31][32].