Objects and files and blocks ... oh my!

I just finished reading an intriguing blog post by Alex Williams at Silicon Angle, responding to an earlier blog by Fred Wilson.

The substance of the debate boils down to whether or not the concept of a "file" (as opposed to a directly-addressed object) will be relevant as more and more computing (and more of our daily lives) moves to the cloud.

Those of us of a certain age may remember taking pride in lovingly assembled CD collections, bookshelves, and video libraries. Of course, these are all being replaced by digital files on our iPhones or tablets.  And, it is reasonable to suspect that those locally-stored assets will soon be replaced by their equivalent in an always-available cloud.

That having been said, while the methods of delivering storage may change, the fundamental requirements of storage are less likely to change. These include capacity (how much data you can store), performance (how quickly you can read and write data), availability (how safe your data is), and compatibility (how easily storage integrates with the rest of the IT environment.) It is this compatibility requirement that is the issue in the file/object debate. As the folks at Silicon Angle point out, 90% of all applications are designed to work with files using POSIX, and all of the surrounding infrastructure is designed to work in a file-based world.

As I discussed in a previous blog post, there is a need for storage that unifies the file and object approaches. I would draw the analogy to the world of automobiles. All electric vehicles may be great, but they require fundamental changes in consumer behavior and fundamental changes in the infrastructure to deliver power to cars (e.g. a network of charging stations in homes, workplaces, garages, etc.). By contrast, hybrid vehicles fit well both with current consumer behavior and with the current network of filling stations. Perhaps, not surprisingly, hybrid vehicle adoption has been a couple of orders of magnitude greater than all electric vehicle adoption. Similarly, I think that hybrid file/object approaches will take hold far faster than approaches which force people to rewrite their applications and otherwise change their IT stack.

Indeed, more generally, I would argue that the era of the specialized storage solution (along with the era of the monolithic, proprietary, hardware-bound storage solution) is coming to an end. As storage is increasingly  consumed as a flexible, scalable, highly distributed pool across public and private clouds, storage itself must embrace the concept of AND-file and object;  on-premise and public cloud; faster, cheaper and better.

Ben Golub is President and CEO at Gluster. He is on Twitter @golubbe.