Join Shailesh Mittal of Datera on Thursday, August 8, 2019 for the Kubernetes Virtual Summit where he’ll be presenting “Overcoming Storage Challenges in Kubernetes” – alongside keynotes from Tim Hockin of Google Cloud, Dan Kohn of CNCF, Abby Kearns of Cloud Foundry, and Liz Rice of Aqua Security. Registration is free but space is limited.
There’s a fundamental misconception that storage is going to slow everything down — migration to containers, the devops cycle, and the application itself. Developers don’t want to deal with storage or wait for the storage admin, and we don’t blame them, because most storage systems are designed for the way IT used to work — monolithic apps that live forever and require massive over-provisioning to make sure they work, and slap a container connector on top to “enable” container-based apps. But that dog don’t hunt, because if your approach to storage doesn’t embrace containers and the new approach to IT that containers promise, you will be ground to a halt by storage. Datera engineers helped define the Container Storage Interface so we get it, but just because your storage is CSI-enabled and drives stateful data doesn’t mean the storage system is built for a dynamic future. Datera exists to break away from the past and drive a dynamic future.
Another container storage myth is that everything about container-based apps is stateless. Application instances may spin up and spin down, but the data harnessed and generated by those apps isn’t stateless — it is stateful. Data has gravity, and optimizing data for the new world of container development is essential.
“Old storage” requires the developer to file a ticket and get a storage admin to physically carve out a volume to attach to their app. The admin will respond in due time with a set of questions — how fast, how secure, how big, what level of fault tolerance, when, etc. And the developer is often challenged about their needs — why so many IOPs or terabytes? Feels like the Department of Motor Vehicles to us at Datera.
That’s why Datera enables the admin to set up different classes of storage based on simple app templates and provide an address to the developer. Think “platinum, gold, silver, bronze” levels of service governing the major attributes of that storage so the developer can move fast. Later, when the developer returns with the news that requirements have changed, the admin either adjusts an attribute of the current class (e.g. make “silver” faster) or changes the app from “silver” to “gold,” and then Datera automatically applies all of those capabilities to the data for that application retrospectively as well as prospectively. One AI-driven policy change can save hundreds of steps (and days!) compared to the old, manual methods.
Look for storage answers that mirror the reasons you are moving to containers. Chances are your organization is moving to containers to get better portability, agility, and velocity, which translates to dynamic apps. Dynamic apps need dynamic storage. Dynamic storage is software-defined not hardware-defined. It’s designed for scale-out not scale up. Instances can be spun up, or retired at the same speed as public clouds, but with much higher performance and flexibility. It makes for an easy start (e.g. carve out a virtual volume, rather than rolling in a new rack), for easy expansion, and for the changing requirements your app will have over its lifecycle — fast today, faster tomorrow; available today, continuously available tomorrow.
The trap that many providers of storage “enabled for containers” will lure you into is that your storage needs to be separate from everything else — another island in your organization’s data operation. That is just the opposite of where you want to head, because that ensures that when your ticket shows up, you are depending on the one person who understands your container storage, and when he or she is out or gone, you are screwed. Storage is infrastructure, and it needs to be part of the overall storage approach, which is why Datera provides a block, software-defined, scale-out storage platform that serves up bare metal, virtualized, and container-based applications at the same time to generate velocity and drive the economic outcomes that all organizations are trying to achieve — while arrays just stand in the way.
Join me at 4:00 p.m. Eastern on Thursday, August 8 for Uncharted Rapids: The Challenge of Storage for Kubernetes as part of the Kubernetes Virtual Summit.