Drive and Rack Scale Storage Architectures
[su_spoiler title=”Disclosure” icon=”plus-circle” anchor=”disclosure”]
This post is based on observations I made while attending Tech Field Day 12. I wasn’t compensated for writing this post, nor was I compensated for attending the event. My travel and incidentals to attend the event were covered by Gestalt IT. For more info, see our Disclosures page.
It’s no secret that big data workloads are introducing storage challenges that are worthy adversaries for even seasoned storage veterans. As storage architectures continue to evolve, it’s interesting to see the different approaches storage startups take to solving problems. In this video I highlighted two storage startups that are tackling storage issues in a very interesting way.
Igneous presented at Tech Field Day 12 and my mind was totally blown when they explained that they’ve created an architecture based on these “nano servers” – which are not to be confused with Microsoft’s Nano Server. As you can see in the picture below, the tiny ARM-based computer attaches to the back of the drive like an interposer. Effectively, the drive is now network addressable and independent from the rest of the drives. In this RatioPerfect architecture, the problems of giant failure domains and controller bottlenecks are a thing of the past. To learn more about how they’re doing this and why it’s valuable, be sure to watch the Tech Field Day videos!
DriveScale also presented at Tech Field Day 12 and I was again surprised and impressed by the way their architecture eliminates so many problems we’ve been struggling with. DriveScale uses dumb JBOD with their DriveScale Adapters (pictured below) to translate SAS to Ethernet. This allows them to easily reconfigure logical computing entities with exactly the right ratio of compute to storage. To learn more about how they’re doing this and why it’s valuable, be sure to watch the Tech Field Day videos!