Joint post with Tony Paikeday from NVIDIA.

NVIDIA与Tony Paikeday的联合帖子。

We’ve discussed the importance of MLOps in a previous post. Once businesses commit to integrating devops practices to their AI projects, what’s the first step that IT teams can take towards successful MLOps?

在上一篇文章中，我们讨论了MLOps的重要性。 一旦企业致力于将devops实践整合到其AI项目中，IT团队可以朝成功的MLOps迈出的第一步？

How can IT teams set themselves up for success?

IT团队如何为成功做好准备？

目标1：支持各种应用 (Goal #1: Support a Range of Applications)

An AI platform doesn’t just need to support TensorFlow — or even just the model development workloads. It needs to provide testing pipelines, versioning, sandbox environments, monitoring, and more.

一个AI平台不仅需要支持TensorFlow，甚至还不需要模型开发工作负载。 它需要提供测试管道，版本控制，沙箱环境，监视等等。

For example, you might start creating Kubernetes clusters for AI workloads. That cluster will run a wide set of applications that need access to a variety of datasets and compute hardware — and likely even a variety of protocols.

例如，您可能开始为AI工作负载创建Kubernetes集群。 该集群将运行大量需要访问各种数据集和计算硬件(甚至可能是各种协议)的应用程序。

目标2：可扩展和自助服务基础架构 (Goal #2: Scalable & Self-service Infrastructure)

Like with any platform hosted by IT and DevOps teams, an AI platform should support application scalability and resiliency. And, optimally, data scientists should have self-serve access to new environments.

与IT和DevOps团队托管的任何平台一样，AI平台应支持应用程序的可伸缩性和弹性。 而且，最好是，数据科学家应该能够自助访问新环境。

→ Without a cohesive plan to support the production pipeline as a unified project, individual application silos often become inefficient, unscalable, and fragile.

→如果没有一个统一的计划来支持作为一个统一项目的生产管道，则各个应用程序孤岛通常会变得效率低下，不可扩展且脆弱。

Step back and ask, “How can we make this set of disparate workloads as easy to manage and to scale as possible?”

退后一步，问：“我们如何使这套分散的工作负载尽可能易于管理和扩展？”

解决方案：合并工作量 (Solution: Consolidate Workloads)

If you’re an IT leader, you have an incredible opportunity. The success of your company’s AI-fueled ambitions requires you to enable developers in a new way.

如果您是IT领导者，那么您将拥有不可思议的机会。 公司AI雄心壮志的成功需要您以新的方式为开发人员提供支持。

Get in front of the productionalization crisis by making architectural choices that will centralize AI infrastructure — consolidating people, process and technology.

通过进行架构选择来集中化AI基础架构，从而解决生产化危机，从而整合人员，流程和技术。

On the storage side, use the same centralized storage underneath all of the applications in the platform. For example, Pure Storage’s FlashBlade is great at handling all different IO patterns and has performant access for both file and object workloads, which means it’s well suited for any of these components.

在存储方面，请在平台中所有应用程序下方使用同一集中式存储。 例如，Pure Storage的FlashBlade擅长处理所有不同的IO模式，并且对文件和对象工作负载均具有高效的访问权限，这意味着它非常适合这些组件中的任何一个。

Likewise, NVIDIA’s DGX A100 brings consolidation to the compute hardware. With DGX A100, NVIDIA consolidated what used to be three separate silos of legacy compute infrastructure, each sized and designed for supporting only one specific workload: training or inference or analytics. DGX A100 supports all of these workflows using just one universal system type.

同样，NVIDIA的DGX A100为计算硬件带来了整合。 借助DGX A100，NVIDIA整合了过去由三个独立的传统计算基础架构组成的孤岛，每个孤岛的大小和设计仅支持一种特定的工作负载：培训，推理或分析。 DGX A100仅使用一种通用系统类型即可支持所有这些工作流程。

Now you have just two building blocks to manage — one for storage and one for compute. This infrastructure simplicity is what lowers the threshold to be able to get models into production; there’s already a place where new workloads can run. With the AIRI reference architecture from Pure Storage and NVIDIA, you can now support the end to end AI lifecycle from development to deployment on one elastic infrastructure.

现在，您只需管理两个构建基块—一个用于存储，另一个用于计算。 这种基础架构的简单性降低了将模型投入生产的门槛。 已经有一个可以运行新工作负载的地方。 借助Pure Storage和NVIDIA的AIRI参考架构，您现在可以支持从开发到在一个弹性基础架构上部署的端到端AI生命周期。

带走 (Takeaway)

By using a flexible yet homogenous infrastructure built on one system type, IT teams have the flexibility to adapt to the demands of the business over time. If you have to go heavy on analytics this week and then pivot to training next week, and then pivot heavy to inference the following week, you can support it all without any physical change to infrastructure. Set a strong foundation for AI projects with compute and storage that flexibly adapts to workload demands as they change over time.

通过使用建立在一种系统类型上的灵活但同质的基础架构，IT团队可以灵活地随时间推移适应业务需求。 如果您本周必须集中精力进行分析，然后在下周进行培训，然后在下周进行推理，则可以在不对基础结构进行任何物理更改的情况下为所有工作提供支持。 通过计算和存储为AI项目打下坚实的基础，随着时间的变化，该项目可以灵活地适应工作负载需求。

Learning more about deploying Pure Storage with NVIDIA for AI: AI Solutions.

了解有关使用NVIDIA针对AI部署Pure Storage的更多信息： AI解决方案 。