Last week I read two interesting pieces of news: Violin has finally filled for the IPO and PureStorage has raised the astonishing amount of $ 150M!
You can find a very good article about it from Chris Evans in his blog and I would like to add something more to his analysis.
SSD market is maturing
It’ clear that something is happening and it’s happening very fast. The storage landscape is evolving very quickly and SSD has become a viable choice for many Tier1 applications.
Costs are continuously going down for many different reasons but the most important are:
- use of tweaked consumer grade technology (MLC instead of SLC)
- improved $/GB cost (due to data reduction techniques)
- improved overall efficiency and durability (due to optimizations on how to deal with the media)
The consequence of the cost shrinking is that more and more enterprises, can take advantage of that!
Technology is maturing and it is available for a broader adoption.
You can easily discover three different segments: PCI (or internal Flash), Hybrid arrays and Full Flash arrays.
These three are also targeted by three different kinds of end users, infrastructure, and application.
PCI/internal is the most radical approach, it has a great potential in next generation scale-out infrastructures. In the most common implementations, local flash is somehow shared between nodes to form a distributed cache or a first storage tier. It’s cheap and scales with the infrastructure. It also has some peculiar design requirements and drawbacks (like faster ethernet connections and allocation of resources on the servers).
PCI and internal SSDs are quickly becoming 100% commodity and they will be directly offered by server vendors. SanDisk’s Flashsoft and PernixData, among others, are showing very interesting solutions here. Hyper-converged and VSA-based infrastructures (Nutanix and Simplivity for example) take full advantage of that too, while others are developing very interesting VSAs or similar products (for example, HP and VMware already have products here).
Hybrid arrays use flash as a first tier or as a cache but they maintain the traditional SAN/NAS concepts. They are faster than traditional arrays, (sometime) cheaper and they offer big capacities thanks to the coupling of flash with SATA drives. Software still plays an important role in improving overall efficiency and space savings. These solutions are easier to adopt and manage (you don’t need to design a new infrastructure!). They are good for general purpose environments as well as for specific applications like VDI. These storage arrays have all the characteristics to be well positioned as primary storage in the small to medium sized companies and for Tier 1.5 (2) applications in large enterprises. Companies like Nimble storage or Tintri are doing pretty well while primary vendors have little to show here.
Full SSD arrays are the best in class in terms of performance and have a much lower latency if compared to traditional shared storage. They are designed to transparently manage new and legacy apps with different workloads at the same time.
Some startups, like Violin memory, has been developing their own hardware (but with few software features) others (like Pure storage) rely on commodity hardware with most of the efforts spent in software development (and features).
They are both good and there are many others that are coming up with interesting things.
Back to the news (and about the $ 150 M)
However, primary vendors are quickly closing the gap: all of them have already bought or developed their 100% flash arrays. Some of these products are ridiculous but others are showing a great potential and are very integrated with the rest of the products line-up.
As history has already taught in the recent past (look at 3PAR’s story for example), even if you have a very good product/technology it’s very hard to win tier 1 storage deals in large enterprises. This kind of customers is very conservative and the power of incumbent vendors is very strong.
Some startups have an advantage at the moment but their time to succeed is very limited.
Things have slightly changed in recent years. If a startup has a good idea it needs to develop it as fast as possible and sell the product immediately and globally. This is why they need huge amounts of money in a considerably short time.
Most primary vendors have overcome their pachydermic-like behavior of the last century, they now know that they need to keep up with the pace.
Startups are forced to compress and accelerate their roadmaps and they need loads of money to do that… If you take a look at companies like Nutanix or Pure Storage, you can easily discover similar paths. If they launch a product and it is well accepted by pundits and end users then they immediately look for more (much more, indeed) money to speed up growth and boost their brand awareness/credibility.