Legacy storage-area network (SAN) architectures are notoriously expensive and complex, leading many enterprises to consider migrating workloads off-premises. However, the public cloud model carries recurring costs that add up over time; with the growing petabytes of data that need to be stored for years, the cloud may exceed the cost of an on-premises storage implementation. Additionally, the public cloud lacks the control and privacy of on-premises infrastructure. Software-defined storage and hyperconverged infrastructure have emerged to address this pain point, but many first-generation solutions do not provide the levels of performance or reliability that mission-critical workloads require.
Who is StorPool Software?
StorPool Storage pinpoints these shortcomings with its software-defined block storage platform. The software is designed for cloud hosting, whether for a large enterprise such as a financial services company that is standing up an on-premises private cloud to host IT services for internal consumption, or for a service provider such as a telco or a cloud service provider that is standing up a cloud data center to host externally-consumed IT services.
StorPool has re-written the entire storage software stack underlying the block disk itself, including on-disk formats, protocols and quorum mechanisms. Its software disaggregates compute and storage capacity resources from underlying hardware, from which it aggregates shared pools of performance and capacity resources. The same physical system may be used for both computation and storage, and new physical systems may be added to the pool in the event that capacity or performance limitations are reached (the architecture is scale-out). Access, monitoring and metadata management functionalities are embedded, so that an additional node is not required for these capabilities. The platform integrates natively with OpenStack, CloudStack, OpenNebula, OnApp and Kubernetes cloud infrastructure stacks, and can connect to other cloud infrastructure stacks and hypervisors including VMware, Microsoft Hyper-V, or bare metal infrastructure, via iSCSI connectivity or via a JSON application programming interface (API).
This software-defined approach enables IT professionals to respond to changing application requirements with greater agility because they can adjust resources more quickly. It also stands to improve utilization of physical systems, because compute and storage are decoupled and do not need to scale in line.
StorPool’s architecture supports both server attached storage (both SATA & SAS) and Non-Volatile Memory Express (NVMe) solid state disks (SSDs), and it enables data to be tiered and migrated to hard disk drives (HDDs) to save costs for storing data that does not require ultra-low latency. In addition to improving flexibility and cost savings, one of StorPool’s core objectives is to facilitate shared storage that performs as fast as, or faster than, local solid-state drives (SSDs). To achieve this end, its software enables the SSD to communicate directly with the host server, bypassing the Linux kernel and its associated overhead, among many other things. According to StorPool, it delivers less than 100 microseconds of latency and more than one million input/output operations (IOPS) per node, as recently demonstrated. Additionally, the distributed nature of the architecture enables data to be read from and written to all drives in the system in parallel.
The scale-out nature of StorPool’s architecture improves reliability because there is no single point of failure. Additionally, multiple replicas of data are written across the cluster, with the user having the ability to set the number of copies that are created. For data integrity, StorPool has written a proprietary checksum mechanism that is applied across the data’s lifecycle, from the client to the SSD, and then back from the SSD to the client. It also retains key data management features, including end-to-end data integrity, thin provisioning, snapshots, clones, quality-of-service (QoS), backup and disaster recovery.
Since it is software-defined, customers have the ability to choose the hardware systems on which their StorPool implementations run. They may deploy the StorPool platform in a hyperconverged implementation, where the client and the storage server coexist on the same node, or they may deploy separate client and storage system nodes. StorPool typically provides recommended configurations, and it has the ability to deliver pre-integrated solutions with a number of its partners. StorPool’s software is licensed on a capacity basis, and it may be purchase pre-paid upfront or on a rolling, monthly basis. Solutions are delivered fully-managed, with StorPool providing 24/7/365 support and proactive monitoring and issue resolution according to service level agreements (SLAs).
Storage environments are notorious for requiring tradeoffs between cost, performance and enterprise-grade data services. As we move into the era of NVMe flash and mission-critical applications that require ultra-low latency, vast amounts of storage capacity, and six nines of availability, these tradeoffs are no longer acceptable. The software-defined storage market is crowded with vendors that are claiming to deliver on all of these areas, which makes it challenging for the storage professional to identify the correct solution to eradicate these tradeoffs for their particular workloads.
StorPool is building credibility and experience in helping to accelerate performance, reduce the total cost of ownership (TCO) and improve the uptime compared to traditional all-flash SAN arrays that are serving typical block use cases including databases and virtual desktop infrastructure (VDI). It has production systems running in customers’ environments for over six years, and a large number of those systems exceed 1PB of SSD capacity.