Wednesday, August 14, 2019

Blueprint: Turn Your Data Center into an Elastic Bare-Metal Cloud

by Denise Shiffman is Chief Product Officer for DriveScale.

What if you could create an automated, elastic, cloud-like experience in your own data center for a fraction of the cost of the public cloud? Today, high performance, data-oriented and containerized applications are commonly deployed on bare-metal which is keeping them on premises. But the hardware deployed is static, costing IT in overprovisioned, underutilized, siloed clusters.

Throughout the evolution of data center IT infrastructure, one thing has remained constant. Once deployed, compute, storage and networking systems remain fixed and inflexible. The move to virtual machines better utilized the resources on the host system they were tied to, but virtual machines didn’t make data center hardware more dynamic or adaptable.

In the era of advanced analytics, machine learning and cloud-native applications, IT needs to find ways to quickly adapt to new workloads and ever-growing data. This has many people talking about software-defined solutions. When software is pulled out of proprietary hardware, whether it’s compute, storage or networking hardware, then flexibility is increased, and costs are reduced. With next-generation, composable infrastructure, software-defined takes on new meaning. For the first time, IT can create and recreate logical hardware through software, making the hardware infrastructure fully programmable. And the benefits are enormous.

Composable Infrastructure can also support the move to more flexible and speedy deployments through DevOps with an automated and dynamic solution integrated with Kubernetes and containers. When deploying data-intensive, scale-out workloads, IT now has the opportunity to shift compute and storage infrastructures away from static, fixed resources. Modern database and application deployments require modern infrastructure driving the emergence of Composable Infrastructure – and it promises to address the exact problems that traditional data centers cannot. In fact, for the first time, using Composable Infrastructure, any data center can become an elastic bare-metal cloud. But what exactly is Composable Infrastructure and how do you implement it?

Elastic and Fully-Automated Infrastructure

Composable Infrastructure begins with disaggregating compute nodes from storage, essentially moving the drives to simple storage systems on a standard Ethernet network. Through a REST API, GUI or template, users choose the instances of compute and the instances of storage required by an application or workload and the cluster of resources is created on the fly ready for application deployment. Similar to the way users chooses instances in the public cloud and the cloud provider stitches that solution together, composable provides the ability to flexibly create, adapt, deploy and redeploy compute and storage resources instantly using pools of heterogeneous, commodity compute, storage and network fabric.

Composable gives you cloud agility and scale, and fundamentally different economics.

  • Eliminate Wasted Spend: With local storage inside the server, fixed configurations of compute and storage resources end up trapped inside the box and left unused. Composable Infrastructure enables the ability to independently scale processing and storage and make adjustments to deployments on the fly. Composable eliminates overprovisioning and stranded resources and enables the acquisition of lower cost hardware.
  • Low Cost, Automated Infrastructure: Providing automated infrastructure on premises, composable enables the flexibility and agility of cloud architectures, and creates independent lifecycles for compute and storage lowering costs and eliminating the noisy neighbors problem in the cloud.
  • Performance and Scale: With today’s high-speed standard Ethernet networks, Composable provides equivalent performance to local drives, while eliminating the need for specialized storage networks. Critical too, composable solutions can scale seamlessly to thousands of nodes while maintaining high performance and high availability.

The Local Storage Conundrum

Drive technology continues to advance with larger drives and with NVMe™ flash. Trapping these drives inside a server limits the ability to gain full utilization of these valuable resources. With machine learning and advanced analytics, storage needs to be shared with an ever-larger number of servers and users need to be able to expand and contract capacity on demand. Composable NVMe puts NVMe on a fabric whether that’s a TCP, RDMA or iSCSI fabric (often referred to as NVMe over fabrics), and user’s gain significant advantages:

  • Elastic storage: By disaggregating compute and storage, NVMe drives or slices of drives can be attached to almost any number of servers. The amount of storage can be expanded or reduced on demand. And a single building block vendor SKU can be used across a wide variety of configurations and use cases eliminating operational complexity. 
  • Increased storage utilization:  Historically, flash utilization has been a significant concern. Composable NVMe over fabrics enables the ability to gain full utilization of the drives and the storage system. Resources from storage systems are allocated to servers in a simple and fully-automated way – and very high IOPS and low-latency comparable to local drives is maintained. 

The Elastic Bare Metal Cloud Data Center

Deploying Kubernetes containerized applications bare metal with Composable Infrastructure enables optimized resource utilization and application, data and hardware availability. The combination of Kubernetes with programmable bare-metal resources turns any data center into a cloud.

Composable data centers eradicate static infrastructure and impose a model where hardware is redefined as a flexible, adaptable set of resources composed and re-composed at will as applications require – making infrastructure as code a reality. Hardware elasticity and cost-efficiencies can be achieved by using disaggregated, heterogeneous building blocks, requiring just a single diskless server SKU and a single eBOD (Ethernet-attached Box of Drives) SKU or JBOD (Just a Box of Drives) SKU to create an enormous array of logical server designs. Failed drives or compute nodes can be replaced through software, and compute and storage are scaled or upgraded independently. And with the ability to quickly and easily determine optimal resource requirements and adapt ratios of resources for deployed applications, composable data centers won’t leave resources stranded or underutilized.

Getting Started with Composable Infrastructure

Composable Infrastructure is built to meet the scale, performance and high availability demands of data-intensive and cloud-native applications while dramatically lowering the cost of deployment. Moving from static to fluid infrastructure may sound like a big jump, but composable doesn’t require a forklift upgrade. Composable Infrastructure can be easily added to a current cluster and used for the expansion of that cluster. It’s a seamless way to get started and to see cost-savings on day one.

Deploying applications in a composable data center will make it easier for IT to meet the needs of the business, while increasing speed to deployment and lowering infrastructure costs. Once you experience the power and control provided by Composable Infrastructure, you’ll wonder how you ever lived without it.

About DriveScale  
DriveScale instantly turns any data center into an elastic bare-metal cloud with on-demand instances of compute, GPU and storage, including native NVMe over Fabrics, to deliver the exact resources a workload needs, and to expand, reduce or replace resources on the fly. With DriveScale, high-performance bare-metal or Kubernetes clusters deploy in seconds for machine learning, advanced analytics and cloud-native applications at a fraction of the cost of the public cloud. www.drivescale.com

See also