"SDS" is as hyped as any category in the industry, and co-opted by storage vendors seeking to hitch a free PR ride. SDS has been used to define all of the following:
-
Anything that's open source
-
Traditional form-factor arrays with a separate license for software
-
All types of object storage including appliance offerings
-
Anything qualifying as hyper-converged including appliance offerings
-
Federation and virtualization solutions that don't even include any storage
But in the end, what were the expectations for SDS
and did the storage industry deliver?
The main customer driver for SDS was the promise of cost reduction. The lure of no longer paying premium prices for expansion arrays from Netapp and EMC was massive. Object storage (incidentally shipping 10-years before anyone mentioned SDS) was a step forward in architecture, but always battled growth pains such as adoption hurdles of proprietary APIs, long and complex deployments, high up-front costs, and the lingering specter of Centera (low performance, challenging integrations, and extreme lock-in).
Then hyper-converged arrived and jumped on that bandwagon, arguing that anything with a software-based design for fault resilience must qualify as SDS, even when delivered as a locked-down appliance. Although hyper-converged servers that brought us a glimpse of compute resource sharing between applications and storage, very little new capability has been introduced.
Whether or not SDS is coming off it's hype-cycle highs, it's clear that SDS solutions today have only made marginal gains in reducing the costs of evaluation, procurement, administration, and end-of-life impact. It still takes weeks to evaluate a product, months to operationalize, and in some cases requires a programmer's skillset to configure and tune.
Where does SDS go from here?
Customers are voting with their feet. They’re going to Amazon.
Amazon provides instant availability, elasticity up (and down!), unlimited scale, and an interface (S3) that’s becoming increasingly standard for unstructured data. It’s frictionless. Such a contrast to the 6-month ordeal to get Ceph working. What if NooBaa can deliver this agile experience for on-premise applications with complete hardware and network independence? This would be the ultimate instantiation of SDS. This is what NooBaa is delivering to customers.
