Tales from the Field
by Woody Hutsell, www.appICU.com
Instead of marketing from afar, I have been selling from the trenches and let me tell you the world looks very different from this view point.
I have a variety of observations from my first 9 months of working closely with IT end-users:
- At least 50% of the IT people I talk to are generally unfamiliar with solid state storage. These 50% are so busy worrying about backups, replication, storage capacity and virtualization that it would take a whole screaming train full of end users before they would care about performance. What they are likely to think they know about SSD is that they are unreliable and don’t have great write performance. I always ask these end users about performance or interest in SSD and usually get fairly blank looks back. Don’t get me wrong, their interest in performance or SSD is no reflection on them just a reflection on their situation. Maybe they don’t need any more performance than they already get from their storage. Maybe performance is so far down their list of concerns as to not matter. Maybe they just can’t budget a big investment in SSD.
- Some high percentage of IT buying is done without any real research. So much for technical marketing. You could write any number of case studies, brochures and white papers and these guys wouldn’t learn about it unless the sales person sitting across from them drops in at just the right time immediately after the aforementioned train full of end-users has started complaining about performance (and the IT guy happens to have budget to spend on something other than backup, storage capacity, replication or virtualization).
- These groups are deploying server virtualizationin mass.
- These groups are standardizing on low cost storage solutions. The rush to standardize is driven by the number one reality affecting many IT shops: they are under staffed and their budgets are constrained. The lack of staffing means that it is hard to get staff trained on multiple products and life is easier if they can manage multiple components from a single interface. The lack of budget means that IT buyers have to make compromises when it comes to storage solutions. Because of item #2 (above), they are reasonably likely to buy storage from their server vendor and often find their way to the bottom of the storage line-up to save money.
You might think these observations would be disheartening, but really I think the story is that SSD is just starting to make its way through to the more mature buyers in the market. Eventually, I believe that all IT storage buyers will be as familiar with and concerned with protecting application performance as they are with capacity and reliability.
A case in point, I have run into at least two customers where the drive to standardize with VMWare and low cost storage is crushing application performance for mission critical applications. The good news for these IT shops is they have low storage costs and an easy to manage environment (because they have one storage vendor and one server virtualization solution). The bad news is that their core business is suffering.
From my limited point of view, standardization is something that the IT guys like and the application owners don’t like. You might assume that I think the IT guys are short-sighted, but no, increasingly I am seeing that they just don’t have a choice; they have to standardize or die under a staggering workload and shrinking budget. Something though has to give. A core business of one of these operations was risk analysis. This company deployed low-cost storage and had virtualized the entire IT environment with VMWare (including the SQLServer database). The entire IT infrastructure ran great for this customer but a mission critical sub-terabyte database was a victim of standardization. The risk managers, whose decisions drove business profitability, were punished every time they did complex analyses by slow application response time. The second business is really a conglomerate of some 50+ departments. These departments were not created equally, however, there were some really profitable big departments and some paper-pushing small departments. To the benefit of some end users and the tremendous detriment of others this business standardized on a middle tier storage solution with generous capacity scalability but not so generous performance scalability. Their premier revenue generating department was suffering with, you won’t believe this, 60 millisecond latencies from storage for their transaction processing system. Yikes. For the non-storage geeks reading this blog, a really fast solid state storage system will return data to the host in well under 1 millisecond. A well-tuned hard disk based RAID array will return data in 5 to 7 milliseconds. A 60 millisecond response time is indicative of a major storage bottleneck. Experiencing a 60 millisecond response time on a single request is no big deal but when this is during a batch process or spread across many concurrent users applications get to be very slow, end-users wait for seconds or batch process take too long to complete resulting in blown batch processing windows.
For now, the story for these two environments is not finished. Once companies head down the standardization trail they are pretty confident and committed. Eventually, the wheels fall off and people begin to realize that it is as bad to standardize on all low cost storage as it is to standardize on all high end storage. Eventually, people realize that IT needs to align to business and not the other way around.
As companies amass larger data stores and the price and options for deploying SSD evolves, SSD solutions will become more common in the data center and a part of each IT manager’s bag of tricks. Zsolt Kerekes, at StorageSearch.com, put it best in his 2010 article “This Way to Petabyte SSD” (http://www.storagesearch.com/ssd-petabyte.html) when he said “The ability to leverage the data harvest will create new added value opportunities in the biggest data use markets – which means that backup will no longer be seen as an overhead cost. Instead archived data will be seen as a potential money making resource or profit center. Following the Google experience – that analyzing more data makes the product derived from that data even better. So more data is good rather than bad. (Even if it’s expensive.)”