When Combining Competing Storage Architectures Drives True Digital Transformation

For the past decade, “digital transformation” has been a key business goal for many — applying technology for improved efficiency, increased revenue and enhanced customer experiences. This goal is not easily achieved. Digital transformation requires rethinking storage infrastructure to manage and access all the additional data it brings.

Digital transformation requires data storage transformation. The volume of unstructured data, including data generated by mobile, AI and IoT, is growing exponentially. And today’s data-driven applications must process higher volumes of data at higher performance and lower latencies.

This new data requires more advanced storage infrastructure that addresses these challenges.

Hyperconverged Infrastructure Solutions

Software-defined hyperconverged infrastructures (HCI) are growing in popularity to expand data center storage and compute needs, according to 451 Research “Voice of the Enterprise” report.

In their recent Q1 2019 release, IDC named VMware, a technology partner of ours, as the market leader in HCI software ownership with over 41% of the market share. Other HCI vendors include Nutanix, HPE and Microsoft. HCI improves efficiency, lowers total cost of ownership (TCO) and provides enterprise-class storage and security, while eliminating the need for expensive storage arrays and special storage networks.

Smaller enterprise customers with predictable storage-compute needs are using HCI technology effectively and don’t need to change their approach. However, for larger enterprises and cloud service providers, HCI can result in underutilized resources by requiring that storage and compute scale in lockstep. This constraint results in companies buying more compute than they need, along with associated software licenses, to keep up with the massive data storage requirements. For many use cases, storage is growing faster than compute, and HCI doesn’t provide an efficient way to scale storage independently.

Disaggregated Storage Solutions

With the standardization of NVMe over fabrics (NVMe-oF), the market is seeing more disaggregated storage solutions. NVMe/TCP drives mass adoption of NVMe-oF by separating compute from storage over standard networks, enabling ease of deployment at scale and lowering TCO. The standard, championed by my company, Facebook, Intel, Dell, Cisco, Micron and other leaders, enables disaggregation of storage from compute, so compute servers can share remote pools of low-latency NVMe SSDs. By scaling storage and compute independently using NVMe/TCP, data centers are maximizing utilization with an easily scalable technology, while delivering latencies equivalent to local storage.

Why disaggregate storage? First, to maximize resource utilization and efficiency, driving down TCO. Now data centers can create clusters with any ratio of storage to compute needed to address specific workload requirements, without stranded capacity. This flexibility can be deployed at scale over ethernet networks, at high performance and using a software-defined model running on commodity servers. Second, once data centers pool the SSDs together in storage servers, one can run data storage virtualization software to get more out of SSDs. Endurance can be enhanced. Latencies can be improved, and data services can be delivered to extract more value out of the SSD investment. Finally, this disaggregation model provides significant operational efficiencies. Compute and storage can now be maintained independently. Containers or VMs can migrate from one compute server to another without having to rebuild direct-attached storage. Also, storage servers can upgrade to the latest Flash technology without impacting any of the compute servers.

While disaggregated storage is up and coming, it does require a rethinking of infrastructure management. Additionally, some disaggregated storage solutions rely on RDMA-based fabrics that are difficult to deploy at scale. IT organizations should deploy solutions that use vanilla TCP/IP with ethernet to keep costs down.

When Hyperconverged Infrastructures Meet Disaggregated Storage Solutions

Surprisingly, competing architectures can work together and build solutions that are better than their stand-alone approaches. With new high-performance solutions derived from NVMe/TCP technology, companies can easily deploy disaggregated storage that works with HCI solutions. The many benefits of combining these two apparently disparate approaches include the ability to:

  • Scale storage and compute independently and efficiently.
  • Share disaggregated storage across multiple HCI clusters.
  • Enhance SSD endurance, enabling the use of new low-cost technologies such as QLC SSDs.
  • Deliver advanced data services to improve efficiency (e.g., Compression), data protection and service availability (e.g., RAID).
  • Deploy easily over standard networks with NVMe/TCP.

Because disaggregated hyperconverged storage solves the resource utilization problem, it will become an important driver in data storage transformation, especially for data-heavy industries such as financial services and cloud service providers. These industries are embracing this approach to gain all the benefits of disaggregated HCI: better utilization, lower TCO and greater flexibility.

Setting Up An HCI-Disaggregated Solution

When marrying HCI with disaggregated storage solutions, it’s critical this combined solution is fully integrated so users can manage both HCI nodes and disaggregated storage through the same framework for simplicity and lower operational costs.

Using the latest protocols, such as NVMe/TCP, ensures end-to-end NVMe performance while relying on existing standard networks.

To minimize TCO, it’s best to use software-defined disaggregated storage that runs on commodity server hardware, as traditional storage appliances use expensive proprietary hardware that will increase the cost per TB by an order of magnitude versus the raw SSDs.

Lowering TCO

Reducing or eliminating stranded storage and compute resources represents significant cost savings. The new software-defined disaggregated storage model combined with HCI provides high-performance storage that scales without requiring changes to clients or network infrastructure. With this approach, storage capacity utilization can increase from 30% or 40% to 70% or 80%. Now infrastructures can achieve optimal scale at the lowest TCO.

According to research conducted by Stanford University, “Server resource requirements for hosting applications with direct-attached Flash and disaggregated Flash [show] that disaggregation can lead to 40% resource savings at the same throughput level. Disaggregating Flash is most beneficial when the ratio of compute to storage requirements varies over time and/or differs widely between applications.”

True digital transformation is only possible when companies improve their storage capabilities and begin to understand the benefits of combining disaggregated storage with hyperconverged infrastructure.

The benefits of this approach include lowered TCO, improved flexibility and improved customer satisfaction. As companies embark on the road to digital transformation, the journey must include data storage transformation if companies are to succeed in this massive business overhaul.

Read original article: https://www.forbes.com/sites/forbestechcouncil/2019/12/26/when-combining-competing-storage-architectures-drives-true-digital-transformation/?sh=24aee1837e89