It has been a few years since the term Software-Defined Storage (SDS) was coined but, to this day, one still needs more than a spreadsheet to keep track of all its incarnations and mutations. The storage hypervisor veterans took little time to wear the new fancy label. Look no further than DataCore that started its operation back in 1998. After over a decade of soul-searching, DataCore is more than happy to assume the throne as the original SDS solution provider. In all fairness, they are. There is no doubt about it. Buy SDS isn’t merely a new identity for the good old storage hypervisor; it is much more. In our view, SDS is nothing more than a means to an end. The Holy Grail is a fully commoditized storage environment in which capacity, performance, and durability are orchestrated by a software layer, entirely disregarding the underlying hardware composition.
Primarily there are three camps in the SDS playground. There are box pushers that sell appliances and there are vendors that offer software-only solutions; and there are legacy array manufacturers that, under market pressure, reluctantly introduce their own version of SDS solutions. Enterprises, who usually value their business integrity and continuity more than they do their IT budget, are more likely to buy into to appliance solutions; datacenters and cloud service providers, on the other hand, tend to be willing to roll up their sleeves and install novel software on commodity hardware to save the last penny. At the end of the day, you either spend money to save money, or spend time to save money. All SDS implementations seem to address a particular market segment, but watch out for the hidden agendas of the solutions offered by some of the major brands. Take Oracle ZFS Storage Appliance for example. Its sole purpose of existence is to assist the performance of the Oracle database. In a lot of cases the term “business continuity” matters more to storage hardware manufacturers than it does to their customers. In other words, they must do something to ensure their own business continuity before they worry about yours. This is not necessarily a bad thing, just something to keep in mind.
Although enterprises love all-in-one appliances and service providers are more DIY savvy, in realty, they all find themselves having a mix of storage arrays from different vendors. SDS, conceptually, should afford the system administrator (and end users) a virtualized layer that masks the underlying hardware and make sure that all member arrays play nice with one another. The pressure is on as enterprises started to collect more analytical data and storing them on expensive disks simply does not make financial sense, not to mention that a good 70% capacity of the enterprise arrays is taken up by data that does not need to be retained there in the first place. Without an SDS controller, scale-out is a nightmare. How can we bring Amazon Web Services level of environment to company private clouds? How does Softbank compete with AWS in Japan if it does not have a cost-effective SDS solution? Curiously, the major payers of the storage industry do not seem to care, other than EMC, whose ViPR 2.0 looks really impressive. Then again, EMC has its own set of mind and most of us would love to see their DNAs mutate one of those days. This one, we’ll have to see.
It is important to note that SDS is much more than virtualizing disk capacity. Performance (IOPS), features (snapshots, rollbacks, replication, and so forth), tiering, application awareness, traffic modeling and adaptive control, all need to be virtualized.
Unfortunately, most can’t see the forest for the trees.