As the enterprise edge expands to include semi-permanent remote workforces, IoT, and a range of applications like AI and M2M, they generate torrents of nonstop data that must be stored indefinitely and be available in near-real-time to users and applications.
Legacy storage architectures are failing to keep up with both data growth and user/application demand. While storage innovation is pushing more workloads into the cloud, many startups have found that the average enterprise is not yet ready for cloud-only storage. Legacy architectures and applications are experiencing extended shelf-lives due to tight IT budgets, and many enterprises still prefer to keep certain workloads on-premises.
As a result, storage startups are finding ways not only to accommodate legacy infrastructures, but also developing storage systems that deliver the latest benefits of cloud-based storage to any deployment model.
The nine startups featured here are developing everything from NVMe disaggregated storage platforms to storage management software for mainframes. Their founding teams earned leadership experience at the likes of Dell EMC, IBM, Juniper, and Pure Storage. Collectively, they have raised more than US$750 million in funding, a figure made all the more impressive by the fact that it includes one startup that has not yet disclosed details about its funding.
Year founded: 2015
Headquarters: Longmont, Colorado
CEO: Tod Earhart, former director of firmware development for Micron Technologies
What they do: Develop flash storage software for cloud data centers.
Burlywood’s flagship FlashOS is storage-controller software. FlashOS analyzes application behavior at the flash-controller level and tunes both SSD performance and features to fit each application. It enables enterprises to deploy whatever underlying storage architecture they wish, including cloud storage, all-flash arrays, and hyperconverged solutions.
Competitors include: Incumbents like Intel, Samsung, and Western Digital, as well as such startups as Apeiron, Qumulo, and Zadara
Customers include: None named.
Why they’re a hot startup to watch: Burlywood is targeting the control plane of storage, developing software that accommodates legacy architectures while paving the way to modern, hyperconverged ones.
The startup has enough early funding to seek out early reference customers, having secured a $10.6M Series A in late 2018. CEO Tod Earhart gained management experience in the industry at Micron and Western Digital.
Year founded: 2016
Headquarters: Santa Clara, California
CEO: Pradeep Sindhu, former served CEO and Chairman of Juniper Networks
What they do: Fungible develops processors, software, and storage clusters that, together, enable enterprises to build hyperscale-like data centers.
The Fungible DPU family are Systems-on-Chip (SoC) processors designed to overcome what Fungible says are the two biggest challenges in scale-out disaggregated data centers: 1) inefficient execution of data-centric computations within server nodes and 2) inefficient interchange of data among nodes.
Fungible’s flagship product, the F1 DPU delivers 800Gbps processing and implements the entire storage, networking, security, and virtualization stack. Fungible’s TrueFabric software, a large-scale IP-over-Ethernet fabric protocol, is integrated into the chip. The company says it provides full cross-sectional bandwidth with low average and tail latency, end-to-end QoS, and congestion-free connectivity.
This enables disaggregation and pooling of all data center resources at scale. TrueFabric is interoperable with TCP/IP over Ethernet, ensuring that a data-center spine-leaf network can be built with standard off-the-shelf Ethernet switches.
The Fungible Storage Cluster (FSC) is a storage platform based on the DPU and composed of scale-out NVMe over Fabrics (NVME-oF) storage nodes. A disaggregated storage platform, the FSC implements a separation of storage control planes and data planes. The platform is designed to enable cloud service providers and enterprises get hyperscale performance out of their data centers.
FSCs are the foundation of the Fungible Data Center (FDC). The FDC is an on-premises, turn-key solution comprised of pre-configured racks of disaggregated compute and storage servers powered by the Fungible DPU and managed by Fungible’s Data Center Composer (DCC) software.
Competitors include: Dell EMC, Liqid, Nebulon, NetApp, and Pure Storage
Customers include: Los Alamos National Laboratory
Why they’re a hot startup to watch: Fungible has raised a massive amount of funding–$311M–and they target a similarly massive opportunity, the hyperscale data-center market.
In 2020, Fungible acquired the assets of Cloudistics, a cloud-software platform startup. With a strong leadership team that served in senior executive roles at Juniper, Apple, Pure Storage, and Qualcomm, among others, Fungible is in a strong position to compete in the hyperscale cloud-service-provider market and open up the adjacent on-premises enterprise market.
Year founded: 2014
Headquarters: Kfar Saba, Israel
CEO: Eran Kirzner, who formerly served as VP Software and Solutions, Enterprise Storage Division for PMC-Sierra
What they do: Develop NVMe over TCP software-defined block storage.
The latest version of their storage software, LightOS 2.1, is an NVMe/TCP scale-out, disaggregated storage solution designed to perform like local flash.
When installed on commodity servers in large-scale datacenters, LightOS 2.1 automatically optimizes for I/O intensive compute clusters, such as Kafka, Cassandra, MySQL, MongoDB, and time-series databases. Each storage server in the cluster can support up to 64K namespaces and 16K connections.
LightOS 2.1 also supports containerized environments like Kubernetes that require large-scale clusters with persistent and durable storage for rapid node migration, workload rebalancing, or recovery from failure without copying data over the network.
Competitors include: Ceph, Excelero, Fungible, Pavilion Data, and Pure Storage
Customers include: Finanz Informatik Technologies Services, AMPD, and Equinix Metal
Why they’re a hot startup to watch: Lightbits Labs has all of the right ingredients–solid funding, strong senior team, early customer traction–to compete in a tough NVMe market. The startup has raised more than $54M, and has named customers.
The founding team’s pedigree is also a mark in Lightbits’ favor. CEO and co-founder Eran Kirzner gained VP-level experience at PMC-Sierra and Wintegra. Co-founder and Chairman Avigdor Willenz co-founded Annapurna Labs, which was acquired by Amazon in 2015, and co-founder and Chief Scientist Muli Ben-Yehuda served as chief Scientist for software-defined data-center vendor Stratoscale.
Year founded: 2016
Funding: Undisclosed but said to be working on its second round of funding.
Headquarters: San Francisco, California
CEO: Peter Thompson, previously VP of emerging and developing markets for DataCore Software
What they do: LucidLink develops cloud storage.
LucidLink’s Filespaces is a cloud-native file service intended to deliver NAS-like performance from cloud storage that is delivered as a service. Filespaces gives enterprises the ability to run file-based workloads on object storage and Azure Blob.
Designed to address business needs around storing large data sets and accessing them over distance, Filespaces creates a shared global namespace that acts like NAS even though the data is hosted in the cloud. LucidLink says that it is able to do this by reducing the traffic between applications and remote storage, thereby eliminating distance and latency issues. Performance is further optimized by data prefetching based on a proprietary, adaptive algorithm; parallel TCP streams; local write-back caching; and in-line compression.
Rather than relying on synchronization, LucidLink streams data on-demand directly from the cloud as needed by the application. The single source of truth is kept in the cloud. Frequently accessed data is cached locally. Since data streams directly to and from the underlying object store, LucidLink eliminates the need for a gateway. Additionally, LucidLink provides file capabilities like direct read/write access from the cloud, immutable snapshots, user access control, and global file locking.
Competitors include: Panzura and Dropbox
Customers include: Torti Gallas and WebMD
Why they’re a hot startup to watch: Security and performance have been two major sticking points for many enterprises that are considering expanding their consumption of cloud storage. LucidLink’s ability to securely stream cloud data at NAS speeds allows them to offer a unique service that cuts into the $20-billion NAS market.
Recent surveys such as one by the University of Southern California have found that knowledge workers will expect flexibility to work remotely, even after the COVID-19 pandemic recedes. Since LucidLink securely streams data between the cloud and any object store, enterprises can deliver a uniform experience to both on-premises and remote teams, even when working with data-intensive applications. Finally, the startup has a marquee named customer in WebMD.
Year founded: 2016
Headquarters: Tel Aviv, Israel and New York, New York
CEO: Gil Peleg, who previously worked in storage development roles at INFINIDAT and IBM
What they do: Develop cloud storage-management software for mainframes.
According to Model9, most mainframe data is hosted in proprietary storage silos, most often in expensive tapes and virtual-tape libraries. The siloed data is difficult to access, manage, and analyze. In addition, IT professionals with the requisite skills and experience to manage the infrastructure operations are a dying breed.
Model9’s software is designed to unlock mainframe data and open up a variety of options for cloud or on-premises object storage, data management, and integration with BI and analytics tools.
With Model9, enterprises are able to run standard mainframe backup, archive, recovery, and space-management operations. All functions are offloaded to zIIPs (IBM z Integrated Information Processors) using any storage system connected over TCP/IP, including object storage, NAS, and SAN.
Competitors include: IBM, Dell EMC, and Broadcom through its acquisition of CA
Customers include: America First Credit Union, Sirius, France Galop, and RELX Group
Why they’re a hot startup to watch: The mainframe market tends to be overlooked by entrepreneurs spinning up storage startups. Model9’s focus on transitioning mainframes to the cloud may shake up a slowly evolving market dominated by a handful of incumbents.
Even if the mainframe market isn’t as flashy as new niches like NVMe, it’s still massively important to the world’s computing infrastructure, providing the core computing platform of many of the world’s largest and most critical businesses including banks, insurance carriers, and retail, health and government organizations.
Model9 has raised sufficient funding to compete for a solid foothold in the market, and it has already lined up several named customers in varied verticals.
Year founded: 2014
Headquarters: Cambridge, New Zealand
CEO: Herb Hunt, former GM at C3 AI
What they do: Develop storage controllers.
Nyriad software enables enterprises and cloud data-center operators to replace RAID controllers with graphics processing units (GPUs) for all Linux storage applications. Thus, GPUs are tasked as both I/O controllers and compute accelerators, a technique that minimizes data movement during the processing of large data sets.
Competitors include: NVIDIA, Qumulo, VAST Data, and Zadara
Customers include: Nyriad has yet to name customers.
Why they’re a hot startup to watch: The software behind Nyriad’s storage solution was originally developed as part of New Zealand’s effort to host the large radio-telescope array Square Kilometre Array (SKA). New Zealand lost out to South Africa and Australia, yet key technical innovations of the project live on, including some incorporated in Nyriad.
In late 2019, Nyriad shook up its management team, raised $11M in Series B funding, and brought in serial entrepreneur and investor Guy Haddleton to reshape the business. Nyriad has since launched an early adopter program, with named partners and participants that include HPC Systems, NEC, and Supermicro.
Year founded: 2014
Headquarters: San Jose, California
CEO: Gurpreet Singh, previously VP of product management for Pure Storage
What they do: Pavilion Data’s storage platform consists of two components: 1) control software, HyperOS and 2) NVMe storage, HyperParallel Flash Array.
Each HyperParallel Flash Array delivers performance of up to 120GB/s read and 90GB/s write. The global namespace capabilities of HyperOS’ clustered, distributed file system expand storage capabilities across any number of flash arrays.
HyperOS supports block, file, and object protocols to run natively on any combination of controllers, across any number of its HyperParallel Flash Arrays. Each Flash Array supports up to 20 independent controllers, each of which runs its own instance of HyperOS. Unlike traditional arrays, that enable multiprotocol by running one protocol on top of another, HyperOS runs each protocol (iSCSI, NVMe-oF, NFS, S3) natively.
Competitors include: Apeiron, Excelero, Fungible, Lightbits Labs, and Pure Storage
Customers include: Sony Innovation Studios, Pixit Media, and Holley Performance Products
Why they’re a hot startup to watch: Pavilion Data targets key trouble spots created by the explosion of data associated with new technologies such as AI and edge computing. The startup has secured $58M in funding and locked down impressive named customers. The startup is also led by a senior team with experience at Pure Storage, Veritas, SeaMicro (acquired by AMD), and Hitachi Data Systems, among others.
Year founded: 2017
Headquarters: Abba Hillel, Israel
CEO: Uri Beitler, formerly head of Samsung’s SSD Controller Development Center in Israel
What they do: Develop storage optimization solutions.
The startup’s flagship product, Pliops Storage Processor (PSP), is a key-value-based storage engine in a half-height, half-length PCIe card that can be deployed to accelerate a range of workloads.
PSP accelerates inefficient software functions to optimize management of data persistence and indexing tasks for transactional databases, real-time analytics, edge applications, and software-defined storage.
According to Pliops, while software-based key-value (KV) store operations perform core functions for nearly every database, software-defined storage, and analytics environment, they cause a combination of read, write, and space-amplification problems. For typical database applications, write amplification can grow up to 40x and read amplification to 100x, consuming network bandwidth and hampering SSD performance, latency, and endurance. In addition, these applications generate space amplification in the range of two to seven times, wasting storage capacity on inefficient internal structures.
This amplification results in large part from the inherent inefficiency of typical storage blocks being uniform in size, no matter the variability of what they contain. Use cases such as transactional/time-series databases, real-time analytics, IoT/edge, and SDS applications have variable storage sizes smaller than the typical storage block size; therefore, all suffer from the inefficiencies caused by this data amplification.
Instead of a single storage-object size, PSP natively manages objects of different sizes, efficiently packing and managing those objects and contiguously mapping them to standard SSD blocks.
PSP cards can be plugged into existing servers and used to eliminate layers of the database, file, block, and storage management accumulated over decades by legacy architectures. With the PSP, enterprises can increase storage performance and capacity without requiring changes to most user-application software.
Competitors include: DataCore Software, Fungible, and Scality
Customers include: Sakura Internet, Japan’s largest web hosting provider
Why they’re a hot startup to watch: Pliops made noise recently with a funding round that pushed its total to $115M. In February 2021 it closed a $65M round led by Koch Disruptive Technologies (KDT) and included Viola Ventures, Intel Capital, SoftBank Ventures Asia, Western Digital, and NVIDIA, among others.
The founding team gained leadership experience at Samsung, M-Systems, and XtremIO, and the startup has already locked down Japan’s largest web hosting provider as a named customer.
Year founded: 2016
Headquarters: New York, New York
CEO: Renen Hallak, previously VP of R&D at Dell EMC
What they do: Develop storage software.
VAST’s Gemini is managed storage software that is sold on commodity hardware. VAST’s disaggregated cluster architecture allows users to scale storage independently of CPUs and capacity only when needed.
VAST makes use of NVMe-over-Fabrics (NVMeOF) to enable commodity datacenter networks to function as scalable storage fabrics that combine the performance of NVMe DAS with the efficiency of shared storage infrastructures.
VAST allows storage servers to be loosely coupled in the namespace, and gives them equal access to shared, persistent NVMe devices over NVMeOF. This architecture enables VAST managed servers to be stateless machines that do not have to coordinate I/O requests with one another. By eliminating cluster cross-talk, VAST contends that its architecture scales better and is more resilient than legacy scale-out architectures.
Competitors include: Dell EMC, Pure Storage, NetApp, and WekaIO
Customers include: Squarepoint, AHEAD, and Enterprise Strategy Group
Why they’re a hot startup to watch: In April 2020, a year after launching its first product, Universal Storage, VAST Data raised $100M in a Series C round of funding at a $1.2B valuation. It has now raised a total of $180M.
VAST Data has also recently shifted its business model to take itself out of the hardware supply chain. Since VAST no longer carries inventory costs, customers no longer pay its markup on the white-label hardware and can scale storage capacity at the same unit price for expansions as for large initial deployments.