Some recent acquisitions in the SDS market - Nutanix bought PernixData and Red Hat bought Permabit - highlight a cautionary adage I often heard when working with venture capitalists in the past. When evaluating the future prospects of a funding opportunity, VCs want to understand whether a new business idea is a standalone product or is really just a feature that will quickly be integrated into a platform (presumably owned and shipped by someone else).
Software-defined storage (SDS) is a high growth area that is bringing some strong benefits – better agility, easier storage management, and reduced CAPEX – to those IT organizations that have the requisite skill to deploy it effectively. Some recent acquisitions in the SDS market – Nutanix bought PernixData and Red Hat bought Permabit – highlight a cautionary adage I often heard when working with venture capitalists in my past that is particularly relevant for software products. When evaluating the future prospects of a funding opportunity, VCs want to understand whether a new business idea is a standalone product or is really just a feature that will quickly be integrated into a platform (presumably owned and shipped by someone else). This has less to do with whether the idea is a good one or not, and more to do with understanding the market value of the innovation and what delivery model (how the product is delivered to customers) will maximize that market value.
In enterprise storage, we've seen a number of examples of innovations around which companies were built which then later just became features of most persistent storage platforms targeted for use in the datacenter. 3PAR introduced the concept of thin provisioning, and grew to be around a $200M company before HPE (who at the time was just HP) bought them. Thin provisioning was a great idea though, and today pretty much every enterprise storage platform offers it, not just HPE.
Data Domain introduced in-line data reduction in a secondary storage platform, leading to significantly more efficient use of raw storage capacity in backup and disaster recovery environments. Data Domain grew to be almost a $300M company before EMC bought them. Storwize, a startup that introduced in-line data reduction (primarily based around compression algorithms) targeted for use with latency-sensitive primary storage environments was snapped up by IBM. Today, almost every All Flash Array (AFA) targeted for use with general purpose mixed workloads includes in-line data reduction driven by not only compression and deduplication, but other in-line storage efficiency technologies like pattern recognition, write optimization, the use of delta differentials for both snapshots and replication, and of course thin provisioning.
Similar evolution occurred with continuous data protection and to a certain extent replication appliances (although in the replication space a niche segment, heterogeneous replication, still supports companies like Zerto). Although CDP is not as widely available across enterprise storage platforms, there are no longer any "CDP companies" and replication is a feature that is supported in literally every persistent storage platform (regardless of whether it is sold as an appliance or just as software).
PernixData offered an innovative caching solution (FVP) built specifically for VMware environments as well as a performance-oriented analytics tool (Architect). Because FVP could significantly reduce the amount of external storage needed, it offered significant value for end users. But caching products are a classic example of "feature" not "platform" products, and in my opinion it was inevitable that they would be scooped up by a platform vendor. Nutanix, a leading hyperconverged platform vendor, did in fact buy them, and will no doubt benefit significantly by applying the PernixData technology to their environment. But the acquisition price was quite low (reportedly under $10M) because the PernixData delivery model did not maximize the market value of the technology. If caching is needed for a particular workload, most customers want it integrated into their persistent storage platform. There are some exceptions - flash caching appliances - but the market for those devices has shrunk, not grown, in recent years as caching capabilities have been built into persistent storage systems that feature flash (and maybe also HDDs) as persistent storage.
Permabit was probably best known for their software-based storage efficiency technologies, primarily compression, deduplication and thin provisioning, that ran on heterogeneous hardware. Instead of selling directly to end users, Permabit sold primarily to OEMs and other technology partners that built their software into their own persistent storage platforms, rebranding the technology in the process. End users generally did not want to buy this capability separately – they looked to their enterprise storage providers to deliver it as a bundled feature in their platforms. Permabit actually had a surprising number of storage platforms using their products over the years, and their technology is quite good. In the wake of the Data Domain and Storwize acquisitions, in-line data reduction has become another classic example of a "feature" rather than a "platform". Permabit had the advantage of being able to run on heterogeneous platforms, but in my opinion it was inevitable that someone, most likely an SDS vendor that did not yet offer their own native data reduction, would buy them. Permabit ran into some headwinds in trying to attract new customers partially because of concerns about access to new versions of their software that would continue to support many different storage systems in the wake of an acquisition. There were examples of Permabit customers over the years that used Permabit to get a time to market advantage for data reduction on their platforms, and then later replaced that with their own home-grown data reduction technologies.
The Permabit technology has found a good home with Red Hat's Enterprise Linux-based software stack, and its data reduction capabilities will strengthen the Red Hat offerings. But data reduction technology is a baseline requirement for persistent storage platforms of any and every kind, so this doesn't give Red Hat an advantage against potential competitors, it just helps make a more level playing field for Red Hat. Red Hat sells an SDS stack built around commercial open source technology running on commoditized webscale hardware, so Permabit's ability to support commodity hardware was important for its inclusion into this stack. Permabit was founded in 2000, well before data reduction was a standard feature in storage platforms, but I must say that I am surprised that they lasted this long, given how the evolution of in-line data reduction technology proceeded.
None of this means that startup teams shouldn't continue to come up with innovative ideas whose market value is maximized on systems platforms rather than as standalone products, but it does mean that founding teams, prospective employees, and VCs should set their expectations about exit scenarios appropriately. This is a market dynamic that is a particular concern in the SDS market of which end users should be aware. This issue is discussed in more detail in Successful SDS Deployments Demand Requisite Expertise on the Part of Customers (IDC, August 2017).