Fusion-IO priced its Initial Public Offering last week at $19 per share and immediately traded up to $25 per share, reflecting nearly a $2B market capitalization. The outcome was terrific, both for the company and its investors (who happened to pour in well over $100M into the company), but more importantly the transaction had deeper consequences in the context of its implications for the storage industry.
Fusion-IO’s IPO was the first pureplay storage new issue since 2007 when Compellent, Data Domain, Netezza, 3Par, Isilon and CommVault all went public (and in the years since have all, less CommVault, been acquired for an average of nearly 10X trailing revenues). What is particular and special about Fusion-IO is that it is the bellwether company (even if only in perception) for a new class of storage based on Solid State Disk (SSD) technology, and in my mind marks a significant shift in how new storage technologies will be validated and ultimately valued.
While in the last several decades we have seen tremendous gains in chip technologies enabling rapid provisioning of massive amounts of computing (Moore’s Law) and in networking technologies enabling incredibly fast transfer of data (10, 40, 100+ GbE), the underlying storage backbone supporting these new infrastructure advances has stayed largely the same. Since 1973, when IBM introduced the 3340 Winchester disk drive, storage technology gains have for the most part plateaued. Sure innovations within and around the HDD software stack (RAID, high availability, replication, thin provisioning, dedup, etc.) have optimized scale, performance and reliability of these systems, but the underlying, core piece of technology – the hard disk drive – has remained the same for more than 30 years.
Until recently, storage issues have been viewed as problems associated with scale and the classic response was to throw more and more disk to deal with the data deluge. Regarding performance of applications, the bottleneck was either the CPU or the network.
Things have changed. Today, we are living in a world where applications require massive amounts short-term compute power and storage. The centralization of computing resources via virtualization and the cloud has correspondingly pushed primary storage systems to their limit as the performance (I/O) bottleneck has shifted. Enterprises are now experiencing the effects of disk contention and latency, resulting in significant deficiencies in database and virtualization performance. Traditional “high performance” storage systems, whose spindles and rigid rotating platters simply cannot keep up with the demands of today’s I/O-intensive Web and enterprise applications, are quickly becoming legacy systems. IOPS (read/writes per second) now matter more than ever.
Enter SSDs, whose I/O potential over traditional rotating magnetic disks is on the order of 2 to 3 magnitudes. Unlike HDDs, SSDs retain
in non-volatile memory chips (often in the form of NAND Flash) and contain no moving parts. More importantly, technical advances in SSD management software have resolved issues pertaining endurance, write problems and data leakage appear and the cost equation from a Gig / $ has become more favorable.
So what is the state of play in the storage industry? Clearly companies have awakened to the IO bottleneck and more companies are incorporating SSDs in some form or fashion into their systems design. Companies like Fusion-IO develop SSD solutions in a drive form-factor that plug in directly to servers via the PCI Express bus (not a viable long term solution in my mind, but that’s a topic for a different blog post). Still others such as Avere, GridIron Systems and Nimble Storage have opted for a tiered or hybrid storage approach which relies on software to identify I/O intensive workloads and then run those through an SSD cache. Finally, there is a new breed of storage upstarts emerging who believe that SSDs will ultimately replace HDDs entirely for Tier 0 / Tier 1 primary storage applications and are creating storage subsystems comprised entirely of SSDs. Pure Storage, which just raised an additional $28M in capital, is supposedly building an “EMC killer” out of flash storage – though now word on the street is that this is more of a file system with tiers of SSDs.
Among this final group, is a company that we at RRE invested in called WhipTail Technologies. WhipTail has created SAN appliance made entirely of MLC-based NAND Flash SSD drives and is the acknowledged leader in extremely high performance, pure Solid State Storage appliances. Today, the company sells its appliance into server and desktop virtualization, database acceleration, file sharing and exchange applications where the I/O squeeze is the most painful. Suffice to say, they have no shortage of interested potential customers.
SSD systems are not going to rip and replace existing storage infrastructure overnight. It’s going to be a long, expensive process that will require the re-assembling to the storage software stack is resolve issues of data leakage, security and reliability. With that said, however, I believe we are in the early days of a transformational boom in the storage industry.