Systems vendors do a spectacular job of giving you a wide array of server and storage system SKUs to choose from. Servers can be configured-to-order with variables for just about every component.
With all of this choice, why do you end up feeling trapped by whichever decision you make?
The problem is that today your infrastructure is defined and designed at purchase-time, where you define processing and storage capacity requirements and purchase servers for new applications or workloads before you run them. This might be okay for VDI (virtual desktop infrastructure) or unified communications applications where estimating size and growth is tied to growth in number of users.
It’s a whole different ball game if you’re deploying big data, machine learning or an object store where capacity requirements can be impacted by a wide range of variables. For these data-intensive workloads, buying hardware before you deploy means overprovisioning to be on the safe side.
And once you purchase a server with drives inside, commonly called a DAS (direct-attached storage) server, you’re stuck with that decision. There’s no way to change the compute or storage hardware or utilize the unused compute or storage capacity that ends up stranded in the box (and across clusters). You are trapped inside the box.
Suddenly, all of those server configuration choices on the price list just seem to add complexity and cause you to overspend. And as your data and workloads keep growing to meet the requests from the business, the impact of pre-configured, trapped resources inside DAS servers is amplified across underutilized, siloed clusters.
There’s another way to deploy infrastructure that creates unbounded flexibility and adaptability, lowers your costs, and maintains performance and high availability. This next-gen data center infrastructure is called ‘Composable’.
Composable from DriveScale enables you to create infrastructure from a few primary building blocks: one building block diskless server SKU and one Ethernet attached Box of Drives (eBOD) storage system. The building blocks are purchased from the vendor of your choice and connected to a standard Ethernet network. And with that, you can compose the widest possible array of logical servers – for example, one server with a ratio of one compute node to 3 drives and another with one compute node to 300 drives. Each server is optimized for the demands of the workload and easily adapted on demand as the workload changes.
You can mount a mix of HDDs and SSDs to the same compute node. And you can run multiple applications in a cluster with any variety of logical servers, each optimized to the workload running on it.
Best of all, with a small setup for benchmarking, you can test your application or workload to see how much processing and storage capacity it is actually using and adjust the capacity of either with the click of a button. That way you don’t purchase any more hardware than you need. We call this just-in-time capacity planning.
With DriveScale, you’re no longer trapped inside the box. We give you the controls to create your own logical servers on the fly, giving you the power to eliminate stranded capital and wasted spend. Read our new whitepaper “Get Off Your DAS” to learn more.