One of the most important changes in the storage world is that the last group of startups that "made it" had huge exits. 3PAR, Data Domain and Isilon all went public and then were acquired for over $2 billion each. Compellent and EqualLogic were in the billion range. And even BlueArc did well with about a $600 million buy out. Not too shabby. And there were a number of smaller storage acquisitions with very attractive valuations compared to their revenue traction.
So what does this mean? Well it is important for the end user because they now buy these innovative solutions from market leaders (in some cases this good and in some cases it sucks). It has also changed the value of the next crop of storage startups, many of whom are getting extremely attractive valuations from investors. Hey its only money and may be worth the risk if it leads to a billion dollar exit.
But are there still new categories and innovations to justify the next, next generation?
One of the biggest areas of VC investment and new startups is for SSD storage system startups. On the surface you might just shrug your shoulders and conclude that the traditional storage system guys can just add SSD disk drives and match the value proposition of the startups. That is why it is essential that the SSD storage system vendors do more than just provide SSD and a menu of features. Instead they need to leverage SSD in order to provide value that the other storage vendors cannot.
What might this include? High performance is of course essential but it must also include a compelling price/performance metric. For example the price per IOPS should be less than $1.00. Most traditional storage systems with disk drives are far greater than this - $3.00; $5.00 and higher.
Capacity optimization is extremely important as well. SSD solutions need to drive down capacity costs so it is closer to spinning media in price per GB. Technologies such as thin provisioning are important but are pretty common amongst HDD-based storage systems. Primary deduplication plays a major role here. There is some controversy with using primary dedupe and impacting performance and even data integrity, so naturally it will be essential that both of these issues are addressed. Now keep in mind any new storage technology typically had these same controversies (e.g. differential snapshots and thin provisioning) but eventually these features became mainstream.
I am a big fan of tiering for SSD. Now the pure SSD players have a disadvantage in this regard. But data doesn't require high performance for its full life-cycle. Therefore having some way to move the data to a lower storage tier is vital. Now interpret this anyway you want. Just figure out a lower tier whether its HDD, a third-party storage system, the cloud or a dedupe tier but you have to make the economics work. It all comes down to money.
Quality of Service (QoS) is another capability that makes sense and can be combined with tiering. If you can assign performance SLAs on a per application basis then you can leverage SSD far more effectively. It is about using the resource and not wasting it. Again, this concept is not new but has not truly been embraced by the market. QoS + SSD can create a very visible value proposition that traditional storage systems can't match.
Cloud Storage - A Wide Open Landscape
I love it when the hype cycle is at the stage that cloud storage is presently. It is simultaneously over-hyped and under-hyped. Just when real solutions are making traction, it is losing its sex appeal.
Cloud Storage is real and is making incredible progress. And there are lots of approaches today. There is pure cloud storage solutions. There are cloud storage solutions that focus on unstructured data; others on transactional block storage; application-specific services and solutions; services that focus on individual users; consumer oriented solutions; etc.
There are medium-sized companies that have hundreds of TBs today. That means that in a year or two they will have over a PB. And of course there are large enterprises that have 10, 20 or more PB today. So we are now in the PB storage era. This is a massive amount of capacity that has its own set of challenges just based on sheer size. Storing all of this data on a single logical storage system is easier said that done. How do you maintain any reasonable performance and address the challenges of data integrity that come with massive scale? How do you protect it all? Where does it live? How much space and power does it require? And how the heck will you ever migrate all of that data between systems? And let's not forget cost - which always is a factor - especially with IT budgets being flattened or reduced.
I was just at an event and the question was raised as to whether people were going to be using the cloud to store big data. I suggested that you first need to define big data. Big data means different things to different people just like "data" in of itself means different things depending on who you are talking to. To a movie studio data means video and audio. To a bank data means financial information. And on and on. Big Data is as varied as data. The importance of big data is not the adjective "Big" or the noun "Data" but the verb "Analysis". It is how we get the use out of that data that matters. Like most things in IT this is not really a revelation but a point of clarity that we are making. We have been analyzing certain data forever but now we need to turn to other forms of data (and there is lots and lots of it) in order to get better use out of it. Analyzing data can lead to a new product break through; improving services; streamlining finances; uncovering business risks; etc. Most importantly, it can make IT far more strategic than it is today. And that is a change that is long overdue.
So how does storage fit into all of this? There are vendors that have put together database and storage systems into a single solution in order to optimize performance to run analytics. There are storage systems that have been optimized for performance for database analytics (e.g. SSD storage systems). But Big Data analytics is still emerging and could stumble and fall. Because the value proposition has to be tangible and easily consumed. I know of very few big data analytic projects that have been landscape changing beyond improving query response times.
What needs to happen (and the process has sort of begun) is the convergence of worlds that are otherwise disparate. You need to bring the business analytic world together with the IT infrastructure world and make something that is best-in-class in both.
The constant and relentless trend has been that unstructured data far outstrips any other data type in capacity growth and this reality is perpetual. Yet most of the new storage startups are just focused on SAN-based solutions. Where are the next generation NAS guys? If you are out there I hope you are focused on more than just a new scale out file system or architecture. Because its not that important. If you want the recipe for success - BIG SUCCESS - I have it. Give me a call :)
Data Protection and Recovery
Not only does backup suck but so does replication. This space needs total reinvention but here is a catch - it isn't friggen easy. It is friggen hard. The challenge is not just technological but also with business execution. You must overcome incumbency. You must be able to deal with educating a legacy mindset with customers. You most ensure that you don't break anything as you attempt to fix everything. Additionally, there has been no uber-big wins in this arena. But whomever can solve this problem and can execute in the market can be the next billion-dollar baby.
Note - I purposely left out vendor names in this blog entry.
No feedback yet
Leave a comment
|« Coraid: Why They Matter||Nirvanix + IBM + Cerner = Enterprise Cloud Storage Validated »|