Hammerspace, Inc. has raised $56.7 million in, surprisingly, what is the company’s first VC round of funding. Why has this move happened now, and what issues does the Data Orchestration System address?
Hammerspace, Inc. was founded in 2018 with a focus on making data globally available across disparate platforms. The company markets a solution known as the Data Orchestration System, a global data environment that supports a range of protocols including, NFS, pNFS and SMB. We have three podcasts that introduce the technology and explain why a distributed file system is so important.
- #152 – Global File System Concepts
- #156 – Introduction to Hammerspace (sponsored)
- Data Unpacked 002 – Data Management in the Hybrid Cloud
Until now, Hammerspace has been funded by CEO David Flynn and other private investors.
As part of the news press release, Hammerspace highlighted a desire to take a different approach with VC funding, choosing companies interested in a long-term partnership rather than the standard model that sees investors liquidating their holdings soon after IPO.
This position isn’t a surprise. Hammerspace CEO David Flynn was previously CEO at Fusion-io, the pioneer of solid-state add-in-cards (AICs). That company floated in 2011 at $83/share (above the expected $19/share) but was eventually acquired by SanDisk in 2014 for $11.25 a share (Dell did bid $33/share for the company at one point). Fusion-io wasn’t bad technology; in fact, it kickstarted an all-flash revolution based on add-in-cards and eventually PCIe cards using NVMe. However, Fusion-io had a dependency on two big clients (Apple and Facebook), while the development of the NVMe standard was some years away.
Between Fusion-io and Hammerspace, Flynn founded Primary Data, a first attempt at the data virtualisation technology developed by Hammerspace. You can read more in this post we wrote at the time (and the follow-on links to Storage Field Day 7).
Before getting into Hammerspace in more detail, we should review the current hybrid cloud storage market and see why the Data Orchestration System is critical.
Across both on-premises and public cloud, block storage is tightly coupled with virtual instances, making it difficult to replicate or move data on those devices. This position is becoming increasingly more integrated as cloud vendors use local NVMe devices to improve performance.
Object storage is, by design, ubiquitous, through the use of web-based protocols. Generally (not always), the I/O profile for objects is sequential rather than random. Object storage isn’t designed for sub-file access, so not the most practical for hosting platforms such as databases.
File storage sits somewhere in between block and object. Block storage is almost always formatted with a file system, which holds the structure (metadata) of the content being stored. Protocols such as NFS offer file-based access over the network, while mature solutions are already available to replicate data between disparate geographic locations. The pNFS protocol enables this process to be performed with data that appears to reside across multiple locations at the same time.
Although it continues to be the primary protocol for container-based applications, we don’t see block being the long-term solution for hybrid environments. This is because a file system can provide massive value-add over block and object storage. This potential could even lead to better solutions to combat ransomware, as we will discuss in a moment.
The single most significant advantage of using a platform that presents itself as a distributed file system is control over metadata. Metadata describes the physical content being stored but also enables that description to be abstracted and virtualised. With this capability, we can now improve data mobility, with a direct saving in costs, time and effort.
The metadata within a file system can be extensible and store additional information on where the data can be placed, including geographic and regulatory restrictions on movement. Metadata can indicate how data should be protected and what performance characteristics should be applied. All of these attributes can be delivered autonomously by the abstraction layer, without human intervention. This is how the Hammerspace Data Orchestration System works.
It’s also possible to build encryption into a file system to make the data much more secure and resilient to ransomware. This feature may make solutions like MOVEit unnecessary, as the exposure of a data endpoint becomes a feature of the file system layer.
As we have highlighted, the use of a distributed file system can bring benefits to a wide range of industries that need efficient data workflow and pipelines. However, Hammerspace does require a significant commitment to implement if customers intend to make the Data Orchestration System a critical part of IT operations. It’s arguably easier to simply move data tactically or use migration software (as we discuss in this blog post). As a result, we expect Hammerspace to use the funding for additional customer education as well as to build out a professional services capability to help customers transform to an architecture with the Data Orchestration System at the core.
The Architect’s View®
We believe the Hammerspace approach is the right one for true data mobility. However, implementation of the technology requires a commitment to a particular architectural design. We hope to see more information on customer solutions, including cost savings being made in public cloud fees, as well as operational improvements.
As we discussed in a blog last year, data platforms provide consistency and extensibility. In an IT world that’s becoming increasingly data workflow-centric, it’s hard to see how enterprise organisations will survive in the long term without some kind of data mobility solution at the core of a hybrid cloud strategy.
Copyright (c) 2007-2023 Brookend Limited. No reproduction without permission in part or whole. Post #b4e7.