The DriveScale System
The benefits of modern workloads are significant, but how do businesses deliver data center infrastructure that’s flexible and scalable enough to support growth initiatives? When it comes to big data deployments, most IT leaders make significant compromises, forgoing many of the enterprise capabilities they have taken for granted. Those days are gone. DriveScale is the only solution that enablesIT professionals to take advantage of the unprecedented performance, scalability, and economics of Software Composable Infrastructure.
Software Composable Infrastructure
Software Composable Infrastructure (SCI) is the underlying innovation of the DriveScale System, and it is the key to unlocking the promise of modern workloads for mainstream enterprises. SCI enables organizations to disaggregate compute and storage resources into available pools, and then recombine these resources into clusters, under software control, into any ratio of compute to storage required by modern workloads. SCI provides data center administrators with the efficiency, economics and security, as well as the agility and elasticity of the public cloud, ultimately ensuring that critical resources are no longer trapped in silos and are optimized.
The DriveScale Composability Platform
How the DriveScale System Works
Traditional data center architectures use a mix of commodity servers that are “fixed” in their pre-integrated configurations of compute and storage. This rigid approach often forces companies to overprovision one resource (e.g., CPUs) in order to fulfill the greater need of another resource (e.g., storage). With DriveScale, the ratio of compute to storage in the data center clusters is a software-defined operation using composable infrastructure technology.
DriveScale automatically discovers available physical data center resources; enables disaggregation of servers into separate pools of compute and storage resources, and reassembles servers under software control, and manages and optimizes the cluster.
Administrators deploy an architecture consisting of pools of compute resources (industry standard servers) and storage resources (hard disks in JBODs) and then compose these resources into logical clusters and logical nodes (software defined physical nodes) that are indistinguishable to the software running on the servers from standard physical rack servers in the data center. They can rapidly and easily provision and re-configure physical resource ratios to optimize for different and changing applications and workloads, through software.
The benefits of this architecture for clusters and nodes preserve data locality and local disk performance of traditional rack server deployments. More importantly, administrators achieve responsiveness and remove any constraints around their physical infrastructure.
The economic benefits of being able to quickly respond to changing application requirements using existing infrastructure are considerable. DriveScale customers have seen up to 70% savings with higher compute utilization for multiple workloads and save up to 33% with a reduced footprint, reduced costs on power and cooling with denser CPUs.
The DriveScale System consists of the:
- DriveScale Management Console
- DriveScale Adapters
- DriveScale Software Agent
- DriveScale Central
The DriveScale architectural approach combines the DriveScale System with industry standard servers, JBODs, and 10GbE network infrastructure that is under customer control – either in a private data center or co-location facility, essentially creating a private cloud environment.
In contrast to the pre-integrated storage and compute resources found in the data center, the DriveScale System enables industry standard servers to be disaggregated and deployed as pools of industry standard disk-lite compute servers and storage disks housed within commodity JBODs across the data center.
The DriveScale Management Server and DriveScale Software Agents coordinate to provide the core functionality of the DriveScale System by composing compute and storage components into logical nodes via the DriveScale Adapter and 10GbE network.
To the operating systems and applications running on them, the logical DriveScale nodes are indistinguishable, both in function and performance, from traditional rack servers with local disks.
The DriveScale Adapter chassis is a fully redundant Ethernet-to-SAS adapter that converts commodity JBODs into Ethernet connected JBODs. This 1U appliance houses four DriveScale Adapters, each with two 10GbE SFP+ ports and two 12 Gb SAS connections.
DriveScale Management Software
The DriveScale Management Software provides a graphical interface to easily define, rebalance and monitor compute and storage resources, while an automation-friendly RESTful API enables integration with other data center tools.
The DriveScale System creates clusters of servers composed of the required number of physical drives connected to physical servers. It utilizes a constraint solver that combines user requirements and information collected about the environment to define the right cluster size and configuration.
With the DriveScale System, cluster and server construction becomes a software-defined infrastructure operation that can be controlled and automated at scale. For example, compute resources can be tuned as deployed workloads change; clusters can be upsized or downsized as needed through server and drive rebalancing, and drives can be easily reassigned away from failed servers to avoid the expense for replication.
DriveScale Central provides easy cloud based administration of DriveScale deployments and an off-site resource for observing data center behavior.
DriveScale Central supports a call home capability for deployments, and provides insights about how best to optimize data center environments. DriveScale Central’s own big data engine will use anonymized metadata from myriad environments to provide actionable recommendations for infrastructure optimization. For example, based on patterns gleaned from other customers, DriveScale Central could recommend a rebalancing of resources to avoid an impending capacity constraint; or it could recommend a tactical firmware update to improve performance of a cluster.