When it comes to storage, this new year certainly won’t be boring. A number of technologies and market movements have been percolating over the past few years and will reach full steam in 2018—impacting the landscape for enterprise suppliers and enterprise buyers alike. There is a lot to cover, so let’s dive in. The year of persistent storage class memory (SCM) Storage class memory (SCM)—sometimes also known as “persistent memory,” or PM—provides byte-addressable persistent memory that lives on a system’s memory bus. While applications can treat SCM the same as they do existing system memory, they do so with the added benefit of persistence. It effectively blends SSD drive storage capabilities with memory bus DRAM semantics. In order to fully leverage SCM, the industry must agree on a set of interfaces, and re-architect applications to leverage them. The Storage Networking Industry Association, or SNIA, is the standards body for the storage industry. SNIA has been driving standards around persistent memory for several years, and the standards are beginning to mature.
Intel Corporation has leveraged these emerging standards into a reference implementation for working with persistent storage class memory. Both MicrosoftCorporation and the Linux community have adopted and integrated Intel’s PEM libraries into their SCM offerings.Microsoft began supporting SCM in Windows Server in late 2017 and is shipping test builds of the next version of Windows Server today with more sophisticated support. Additionally, Microsoft now supports the technology in SQL Server 2016 SP1, in order to provide performance enhancements. Oracle Corporation is also supporting flavors of storage class memory.
This is the year that the industry will see in-memory computing, big database, and analytics applications begin to adopt SCM as a core part of their architectures. We will similarly see hardware offerings emerge to support these new applications. It will take longer for mainstream applications to find benefit, but it ultimately will happen.
Optane and 3D XPoint begin to make a difference
Any discussion of storage class memory immediately leads to a discussion of 3D XPoint products, which IntelCorporation is promoting under its Optane brand. Developed jointly by Intel Corporation and Micron Technology, 3D XPoint (pronounced “3D cross-point”), is a flash technology that claims to provide up to 1000 times more endurance, 1000 times more speed, and ten times the storage density of traditional flash architectures.
Optane is still in its early days, but early indications show it living up to its promise. The results published by emergent players like E8 Storage and Apeiron Data for Optane are fairly amazing. Apeiron is claiming speeds 8 times faster than traditional parts, with 38 times less latency. Optane is selling at a premium right now, but will level out as manufacturing ramps up and Micron Technology brings its solution to market later in 2018. 2018 will be the year these parts starts be designed and shipped in systems where fast persistent memory is a requirement. Optane and 3D XPoint will turn up in both traditional SSD and storage class memory. Mainstream adoption, however, will lag into 2019 as costs normalize. HCI grows, but cautiously Hyperconverged Infrastructure (HCI) really requires more attention than can be properly be given in this article, but after several years of availability and evolution from a variety of vendors, we can make some predictions about 2018. HCI will continue to grow at a rapid pace, but largely in the small and medium business space. Broad adoption of HCI in the enterprise data center will remain elusive. The obstacles are less technical in nature and are more a function of IT organizational politics and process. HCI confuses the typical fiefdoms of networking, storage, and compute, and blurs the lines of manageability, procurement and responsibility for a solution's success. However, enterprises will adopt HCI where it makes the most sense, in areas such as branch offices and isolated application deployments.
Who are the winners? Dell Technologies will continue to dominate in the HCI space. Dell has both the broadest and deepest technology portfolio and is providing excellent products with its VCE VxRail and VxRack offerings. Nutanix will remain the number two HCI player. The wild-card around Nutanix is its apparent shift into more of a software-only offering, which will give HCI revenue to a number of OEMs. HPE and Cisco will become second-tier players, having little value to offer over Dell and Nutanix. As for NetAppNTAP +0.6%, it’s a little too early to know. The NetApp HCI solution is unique in that it allows separate scalability for computing and storage, and breaks down some of the barriers to enterprise adoption.
Predictive Analytics become table stakes
No IT administrator really wants a fully automated data center; the trust isn’t there that the machine will make the right decision. In contrast, every IT administrator wants intelligent insights derived from deep analysis of operations data— insights and recommendations that help IT make better decisions. Predictive analytics provides that, and will become table stakes in 2018.NetApp pioneered predictive analytics with its AutoSupport feature a decade ago, which has since evolved into NetApp Active IQ. Using many of the same engineers who developed NetApp’s solution, Nimble Storage delivered the excellent Infosight Product, which HPE executives called a “crown jewel” of the Nimble acquisition. HPE has announced the integration of Infosight and HPE’s 3PAR storage products, and have not been shy trumpeting its ambitions to leverage Infosight across the datacenter. Pure Storage offers predictive analytics that are close to on-par with HPE’s Infosight, as part of its Pure 1 Global Insight Offering. IBM Storage also offers its own predictive analytics. It’s surprising that the vendor most lacking in predictive analytics is Dell Technologies. The EMC team does offer some automated support, but doesn’t seem to have the depth in this area that some of its competitors have. I expect Dellwill find a way to quickly close this gap in 2018. NVMe commoditizes and NVME-F starts to take hold NVMe is a high-speed controller-less interface that connects flash memory to a storage bus, providing nearly six times the performance of traditional flash interconnect technologies. Adoption of NVMe began in earnest in 2017, but 2018 will be the year that nearly every new flash offering will arrive with an NVMe interface. Components are being delivered today from Intel Corporation, Samsung, and Micron Technologies. NVMe is all about connecting flash to local hardware. Its companion specification, NVMe-Fabric (or NVME-f), extends the protocol outside a local storage domain. Implementations of NVMe-F are being delivered in 2018 over high-speed Ethernet, fiber channel, Infiniband and hostless adopted fabrics.
Innovative early adopters are delivering products with NVMe-F today and are achieving impressive results. Pure Storage supports NVMe-F on its DirectFlash shelf. E8 Storage is leveraging NVMe-F in some of its software-defined flash storage configurations. The tier-one OEMs are lagging, with only IBM Corporation publicly demonstrating a product. This will start to change in 2018, but as it typically happens with all-flash arrays, it will be the nimble, more innovative technology companies leading the way.
Flash is about density
While we’re talking about Flash, the continuing story is all about physical density. Recognizing that modern flash doesn’t really need to fit legacy form factors for storage, Intel Corporation shook things up with the introduction of its ruler form factor. The ruler creates a world where it’s possible to populate a whopping petabyte of raw flash into a 1U form factor. We’re seeing OEMs and ODMs both adopting the ruler form factor, and starting to deliver variations to allow more custom packaging.
The top storage OEMs will hold steady
Beyond technology and market trends, there will be some activity in the OEM world. OEM leadership in the storage sector isn’t going to significantly shift in 2018, but the leaders will be challenged by a number of the trends discussed. Dell and Hewlett Packard Enterprise are perhaps the best insulated from disruption, given the breadth of their respective portfolios. Dell , in particular, is poised to execute in the converged and composable infrastructure markets. HPE won’t sit still—it will continue to acquire technology to fill the gaps.
Pure Storage continues its growth trajectory, avoiding acquisition
Pure Storage emerged as a strong player during the early days of all-flash arrays and has leveraged its early success into continued growth. It continues to differentiate against its competitors with good technology, delivered in feature-rich products which are delivered quickly to market. The company’s new CEO, Charles Giancarlo, is backing up strong technology with extremely aggressive vertically-targeted marketing—and it’s working.
There were strong rumors during mid-2017 that Pure Storage had put itself up for sale. Maybe that was true, I have no inside knowledge. The current leadership, however, is not executing like a company who wants to be bought. The reality is that Pure Storage, with its market cap of ~$3.8B, and increasing revenues, would be a very expensive acquisition that only a few tier one players could afford. Cisco Systems is the largest OEM without a compelling storage story, but Pure Storage doesn’t have the breadth of technology or depth of customers that Cisco would look for in a storage buy (and Cisco did pass on Nimble Storage last year). Lenovo wouldn’t have the appetite for an acquisition that large, and HPE won’t be ready to supplant Nimble and 3Par with something newer until at least 2019. This year Pure Storage will continue to execute and innovate, and it won’t be acquired.
Maybe my most surprising prediction for the year to come is what I’m predicting isn’t going to happen. There will be no revolution in cloud. Enterprises and technology providers will continue to struggle for answers on how to best leverage and balance cloud with existing IT, and the world will slowly evolve towards a solution that works. Acquisitions, when they occur, will be small. Object storage is not taking over the world. NAS is not dead. Neither are hard drives or even tape.It’s going to be an interesting, but not a revolutionary year. Storage trends are converging into new solutions and use cases. OEMs are talking more about solutions than they are products. Flash gets faster and denser, and moves onto the memory bus. Convergence happens where it makes sense. OEMs continue to exist in their current pecking order, for now. The storage world continues to evolve.