Enterprise storage demand is growing due to the increasing amount of data generated by IoT, artificial intelligence (AI) and other big data applications. The introduction of advanced wireless networks, such as 5G will increase the need for storage, not only at the data center, but also at the edge and in endpoints. IoT and AI are enabling more efficient factories (Factory 4.0) that can make more goods with greater yields and using less energy and raw materials.
Bruce Kornfield, CMO and GM of Americas for StorMagic says, “AI and ML will spike in 2020 and continue to drive the global data explosion and subsequent need for storage. As both absorb data from a variety of sources to attempt to correlate and identify patterns in distinct or divergent data sets, the rise in implementation of such technologies will contribute to data production as well as data ingestion.”
At the same time, trade-offs between storage costs and latency/performance are driving the growth of all types of storage devices, such as HDDs, SSDs and magnetic tape as well as emerging non-volatile memories that could replace energy hungry SRAM and DRAM. NVMe and NVMe over fabric (NVMe-oF) are enabling today’s highest performance primary storage systems.
Enterprises are using sophisticated software-controlled storage solutions both in local data centers, at the edge and in the cloud and there are new computing architectures that are bringing processing closer to the data storage. In this blog we will explore trends in storage systems that will dominate the industry in 2020.
2019 saw increased roll-out of NVMe-oF storage systems, often using fibre channel (FC) for the fabric. 2020 will see continued use of FC but also more implementation of Ethernet-based NVMe-oF solutions. Some NVMe-oF solutions will be hybrid flash and HDD-based systems, increasingly using three-bit per cell and even four bit per cell NAND for high density flash memory applications.
NVMe provides fast access to storage, but doesn’t address the latency between many storage systems and processing. Various types of local specialized processors, often called accelerators, can speed up the rate of processing. One type of accelerator focuses on processing close to digital storage. This approach is frequently called, Computational Storage and can put specialized processors in the storage devices or in the storage network close to the storage devices (such as products offered by Eideticom). A Computational Storage working group in SNIA is busy working on new standards for the field.
Nader Salessi, the CEO and founder of NGD Systems (which includes a storage processor in its SSDs) says, “Computational Storage, especially the way we marry the use of NVMe SSDs and compute power, adds analytical power and speed so that results can be accomplished right away and where the data is generated.” New wireless networks, like 5G will especially benefit from local processing of high bandwidth data at the edge using computational storage devices. In addition to these solutions Western Digital is planning to introduced programmable processors using RISC-V chips in its storage devices starting in 2020.
The movement of content to the cloud (remote data centers that sell capacity on an as-used basis) continues but it is clear that for many organizations a hybrid cloud that includes on-premises storage as well as public cloud storage will be the dominant approach for some time to come.
Cloud storage company, Object Matrix, points out that most storage users don’t want to worry so much about what their content is stored on. They want storage to just work and be generally invisible, whether located on-premises or in a remote data center.
Matt Kixmoeller, VP of Strategy at Pure Storage, says, “Object Storage has shaken off its roots as cheap-and-deep cold storage and has started to emerge as the new form of primary storage. Object Storage has become the storage standard for cloud-native applications - for its ability to support highly parallel and distributed access to large data sets…Object Storage will become the natural choice for enabling applications to decouple and disaggregate applications and their compute resources from a pool of shared storage. This pattern has taken hold not only in custom SW development but also with large software vendors such as Splunk and Vertica.”
However, the software used in many workflows use file-based storage, requiring some type of file gateway to object storage in a private or public cloud. Companies such as Tiger Technology are making object storage gateway products that support hierarchical storage management (HSM) to quickly recreate files from object storage.
As the information stored in data centers, at the edge and at endpoint increases so will the energy consumption used to store and process this data. The next decade will see major efforts made to control energy consumption including running data centers at higher temperatures and replacing energy consuming volatile memory with non-volatile memory (such as MRAM, RRAM or FRAM). Doing more with less in data processing and storage will lead to realization of the green data center.
2020 will see increasing network storage using NVMe-oF, more computational storage offerings, growth in hybrid and public clouds, increasing use of object storage and object storage gateways and a move to greener data centers.
