As I listen to my esteemed colleagues in the OpenStack world, and the DevOps and Cloud world in general, say that there is no demand for Object Storage, I get all sad. Why? Because that means that they’re mounting volumes. It means they’re mounting volumes and storing their precious there. That means that the cloud platform is most likely doing something complicated and expensive to replicate this data. That means that Amazon’s S3 didn’t really change anything and we’re just developing enterprise software again.
The great limitation of Object Storage is doing seek operations on files. Seeks break down into reads and writes.
Reads include the CDNs aplenty that have solved this problem, but it’s a read-only solution for jumping around in video. I don’t have a problem with that, as it’s an API based service that is easily implemented and accessed.
However, the other big use case is writes, and the culprit is SQL (and some NoSQL) Database Files. Seeks through those files are critical to their operation.
I’m interested in what other requirements are driving the mounting of block devices, and what’s distracting app developers from the great awesomeness of Object Stores like Swift.