Microsoft high-performance computing with Azure
Articles,  Blog

Microsoft high-performance computing with Azure


It’s impossible to look away every sense is captivated pulled into the dance of color unfolding before you. This is the unique immersive experience that customers see when they enter a Microsoft retail store. Over 80 of our welcome customers like this with a huge array of video walls the content displayed on these screens is constantly being updated to highlight new products and campaigns. But updating the content this frequently comes with a lot of challenges. The layout of screens is unique to each store and filling them requires imagery tens of thousands of pixels wide or tall. A ton of computing power is required to work with this scale of visual content. During the production phase, designers need to preview and iterate on their work. When everything is finalized, each video needs to be rendered, split into sections, and encoded for each and every store. And the entire process requires a massive amount of storage for the many high rez assets, images, previews, and final videos. Originally we used an on-premise render farm for this, but it was a ridiculous waiting game up to two days in limbo while the video files were rendering and encoding. And if we needed to change anything, boom clock resets to zero. Our timeline for each new video was a painful 12 to 14 weeks. We needed a flexible, scalable, cloud based solution. One that allowed creatives to actually be creative while delivering to Microsoft’s unique store configuration. We looked into third party solutions but they didn’t satisfy our needs. Then the team had a light bulb moment Azure Batch. With the ability to run design, rendering, and encoding applications in the cloud, it was a perfect fit. So we engineered a workflow with Azure Batch to automate production in rendering and the results are game changing. Batch can duplicate the power of hundreds to thousands of on-premise render nodes with virtual machines. The actual process of rendering is computationally intensive. There are many frames and each frame can take hours to render. With Azure Batch, we can split a job into many pieces, rendering the frames in parallel across multiple VMs and completing the entire process much faster. And with applications like After Effects, Maya and FFMPEG running on the VMs. We can extend the power and scalability of Azure to every stage of production. On the front end, we use Azure to host a web based application that lets designers upload assets, arrange layouts, preview results, and finalize content. Before designers might sacrifice great ideas to avoid a two day render, but with the wait time reduced to as little as 30 minutes, they have the freedom to make revisions on the fly and distribute them faster than ever. Virtual machines are created on demand and spun down when the work is finished. Dashboards enable real time performance monitoring. Hardware specifications can be changed with a couple of clicks, and exabytes of capacity on Azure Blob Storage allow the finished files to stay in the cloud letting us push videos to every store in a matter of hours, not weeks. In total, we cut production time to as little as two weeks, and reduced operating expenses by nearly 75 percent. It really is a revolutionary workflow for anyone producing large scale content. Whatever the application, Azure Batch can make it happen faster than ever before. For us, that means we can focus on creating a stunning customer experience with everything automated in the cloud.

Leave a Reply

Your email address will not be published. Required fields are marked *