Scalable Solutions for the AI Data Pipeline: FlashStack for AI


January 17, 2019 - 1 Comment

This blog was co-written with Francoise Rees, Cisco Solution Marketing manager for AI and ML

Cisco UCS C480 ML M5

Last  September, Cisco announced a new UCS platform, the Cisco UCS C480ML M5 designed specifically for intensive AI and ML workloads. Cisco has met and worked with customers to identify workload patterns that make up individual data pipelines. As Cisco’s Han Yang presented in his blog, no data pipeline is alike and customers developing artificial intelligence and machine learning based applications depend on dedicated platforms to manage their sophisticated and complex data pipeline; an infrastructure for data ingestion, dedicated clusters for workload compute processing, and a storage platform that easily scales and protects the raw and subsequent processed data.

Introducing FlashStack for AI workloads

Cisco and Pure Storage have teamed to provide a solution for deep learning and AI workloads. FlashStack for AI combines the compute scale of the Cisco UCS C480ML M5 with the storage scalability of Pure’s FlashBlade and  packaged within the FlashStack platform. This fully converged infrastructure for AI and ML applications offers:

  • The Cisco UCS C480 ML M5 system – A purpose-built rack server for deep learning and machine learning that includes eight NVIDIA Tesla V100 Tensor Core GPUs with NVLink Interconnect
  • Pure Storage FlashBlade – An elastic scale-out storage platform that delivers performance that scales linearly as data grows.
  • Cisco Nexus switches – They provide the massive network bandwidth required by AI and ML applications

With FlashStack for AI workloads, organizations can deploy a validated architecture for AI workloads that reduces the design risks in building a data, compute, and storage infrastructure for the AI data pipeline and help achieve better business outcomes.

The amount of data that needs to be processed for AI is massive. High performance, architecturally optimized solutions that can harmoniously exist within your existing environment is critical. IT teams need to easily address these new workloads without increasing complexity. FlashStack for AI workloads provides a proven, validated design for these highly data intensive workloads.

This is an extension to the existing and successful FlashStack portfolio for traditional workloads. With FlashStack for AI, existing FlashStack customers can easily adopt AI/ML for their business without re-architecting their data center. Here are just a few use cases for FlashStack for AI :

  1. A 24/7 light out warehouse that uses automated, intelligent robotics ready to meet high demand customer shipment requirements
  2. A virtual concierge that responds to customer online queries and suggest purchases based on customer choices
  3. A more secure banking transaction program that uses intelligent algorithms to detect fraud patterns.

The FlashStack for AI validated designs will be available later this year. Visit Cisco’s validated designs  for FlashStack website to learn about recent CVDs for Cisco ACI, Oracle and SAP TDI.

 



In an effort to keep conversations fresh, Cisco Blogs closes comments after 60 days. Please visit the Cisco Blogs hub page for the latest content.

1 Comments

  1. And there's more! Please read Pure Storage's blog about other exciting announcements from Pure Storage: https://blog.purestorage.com/putting-your-data-first-in-2019-pure-returns-to-cisco-live/