michelangelus - Fotolia
Containers, clouds, NVMe drive 2020 data storage innovation
Containers on bare-metal flash and souped-up processing highlight innovations in storage, aiming for seamless data movement between local and cloud arrays.
Over the past two decades, well-funded startups have driven data storage innovation.
Storage networking (Brocade), data deduplication for disk backup (Data Domain), clustered NAS (Isilon), Ethernet storage (EqualLogic), hyper-converged infrastructure (Nutanix) and all-flash arrays (Pure Storage) became mainstream due to successful startups. In each of these cases -- as well as cloud storage -- a group of newcomers came along around the same time to create or advance a new category. A few became public companies, more got acquired by large vendors, and others went under.
Because of the influx of venture capital and companies popping up, it was easy to identify trends leading to the next big thing in storage.
And now? Most of the funding in the storage world over the past two years went to data protection vendors that can't be classified as startups. These include Veeam Software, Rubrik, Cohesity, Actifio and Druva. These vendors helped advance virtual, converged and cloud backup, and copy data management. But all of them had products already in the market for years before their latest funding hauls and are no longer next-gen vendors.
For primary storage, funding is scant these days with no obvious area for innovation.
That doesn't mean there is no "next big thing" in storage. It likely means changes will be more evolutionary. According to interviews with experts and storage vendors, there is no single technology that appears ready to shake up the industry. They expect the next round of storage breakthroughs will involve assimilating recent advances such as storage containers, NVMe flash fabrics and faster processors into a coherent ecosystem.
Hot areas in storage
Unlike the early days of flash and cloud, the barrier to entry is much higher for new storage companies, said Scott Sinclair, a storage analyst at Enterprise Strategy Group. Sinclair said AI, edge and new media will disrupt storage for the next several years, more than any changes to storage itself.
"If you think about storage, it's very difficult now to differentiate yourself as a startup. We've almost hit a law of diminishing returns, in terms of coming up with a net-new feature that really changes the game," Sinclair said.
Dell Technologies Capital invests about $150 million annually in IT startups. The fund has stakes in about 115 small tech companies in infrastructure, data management and storage.
"One area we're paying close attention to is developer-led infrastructure. With digital transformation, developers want to develop new software in containers, without worrying about the underlying storage," said Gregg Adkin, Dell Technologies Capital's managing director.
Adkin said AI, machine learning and streaming media are changing storage architecture for cloudlike agility. Key investment areas include performance-based object storage, cloud and new storage media. Recent investments include startups NVMe over Fabrics company Lightbits Labs, chipmaker Nuvia and all-flash specialist Vast Data.
Containers now mature, and need storage
In just a few years, containers have matured from an IT niche to a robust tool for faster application development. Thanks largely to the emergence of Kubernetes, containers are poised to disrupt data storage innovation for AI, extreme analytics and virtualization.
Startups Portworx and StorageOS were early to market and more are preparing to launch.
"Containers are getting to be a big deal, specifically Kubernetes container storage. That's where you're going to see a lot of innovation and startups coming out this year," said Marc Staimer, president of Dragon Slayer Consulting.
Hardware-based Kubernetes farms
Running on bare-metal is considered the optimal deployment for containers. Most major storage array vendors have added support for containers on their hardware, and new container-focused hardware platforms are moving to market.
Staimer said vendors are working to bolster security and storage efficiency of containers, with an emphasis on data security. "You're going to see more people deploying them, because you can get about six times as many containers on the same hardware as you can with virtual machines," he said.
Startup Diamanti (formerly Datawise) sells D20 hyper-converged infrastructure appliances for large-scale deployment of Kubernetes clusters. The Seattle-based vendor's latest product is Spektra, a global control plane that accelerates Kubernetes workloads locally and in the cloud.
The Diamanti appliance is an x86 white box with two attached custom PCIe cards. Spektra functions both as RAID controller and smart NIC. The startup claims several Fortune 500 companies as early customers.
Despite having ample IT resources, large companies find the nuances of Kubernetes to be intimidating at first, Diamanti vice president of product Brian Waldon said. That has enabled Diamanti to carve a niche selling its bare-metal system to companies that have already failed in trying to master Kubernetes.
"Typically, we come in after a DIY project has failed. We are the easy button for Kubernetes, with a physical form factor that doesn't rely on virtualization," Waldon said.
Kubernetes is associated mostly with ephemeral workloads, but Waldon said Diamanti's architecture has won customers that rely heavily on transactional databases such as MariaDB, MongoDB and PostgreSQL.
"Because storage is integrated into our product, customers are comfortable running their stateful workloads on our platform," Waldon said.
Weaving storage fabrics
Brian CarmodyCTO, Infinidat
Brian Carmody, CTO of disk array vendor Infinidat, said data fabrics represent the next wave of data storage innovations. That's because these fabrics make it easier to transparently move data between physical and virtual targets. Infinidat Elastic Data Fabric encompasses the vendor's InfiniBox arrays, data protection and consumption-based services. The Infinidat fabric acts similarly to NetApp's Data Fabric for file storage.
"We're at the end of an era in storage. We've perfected the idea of the cloud, and the next big thing is going to be taking place at the edge," Carmody said.
He said traditional storage arrays are a "technological dead end" that will be replaced by "software-defined clouds" that extend the core to the edge.
"You know who's positioned for this? Telecom companies, but they're not going to have a dozen hyperscale data centers in the middle of the country. They're going to have hundreds of thousands of micro-data centers," Carmody said.
Compute moves closer to data
Computational storage is gaining momentum as organizations generate more and more data outside the core data center. Tools for accelerating network performance should drive greater adoption of storage class memory, especially with the arrival of Intel Optane.
Newcomer Pliops is sampling a hardware accelerator for AI workload with cloud providers, server and storage vendors. The Pliops Storage Processor (PLP) offloads work from the CPUs for faster reads and writes over a network, or could be used to enable the use of ARM servers and lower-cost processors.
Pliops use cases include data analytics and transactional databases. The dedicated PCIe architecture runs on most systems. PLP boosts the storage modules that manage repetitive tasks in open source databases.
Pliops chief business officer Steve Fingerhut said system bottlenecks have swung back to the networking side, a dynamic created by the rise of faster flash and storage class memory technologies.
"We are [where] databases and flash storage come together," Fingerhut said.
Computational storage could be a key technology as AI apps mature, said Steve McDowell, a senior analyst of storage and data center technologies at Moor Insights & Strategy. NGD Systems is a smaller vendor developing memory-based offload cards, but Micron Technology and leading storage vendors are also working to develop computational storage devices.
Although it won't hit the mainstream this year, and maybe not for several years, McDowell said computational storage will gain momentum with growth in unstructured data. "This is an area where there always seems to be a constant stream of low-level activity. The idea is to put some compute capability directly next to my flash, to do preprocessing or other activities that enhance the application's performance," McDowell said.
Nuvia, a company launched by three former Apple Corp. developers, also wants to take on computer processing leaders Intel and Advanced Micro Devices. Nuvia is developing processors for cloud data centers. The startup received $53 million from Dell Technologies Capital.
Dell Technologies Capital's Adkin said new data storage technologies will center on pushing memory as close to data as possible.
"Moving these new media and memory technologies closer and closer to the CPU means you get to rewrite your storage stack. You can improve performances and improve capabilities by having it there, and we're at the very early stages of that," Adkin said.