Download the authoritative guide: Enterprise Data Storage 2018: Optimizing Your Storage Infrastructure
I do not think that REST solves all of the world’s data storage problems. And I have ranted enough over the years about the lack of innovation and improvement that have happened to the POSIX standard. But POSIX interfaces for applications still solve problems that REST cannot.
For example, say you have a large dataset that you need to process. With REST you cannot start operating on it until the whole file had been transferred, while with POSIX you can start reading the file within your application.
REST and POSIX interfaces for storage solve two different problems. You do not want to fix your toilet with a hammer nor pound nails with a pipe wrench, and up until REST came on the scene there was only one tool in the toolbox. Now we have two but do we need more?
I think the answer is likely yes, and a good example might be data analysis applications, which are going to need ways of ingesting data in real-time to allow decisions to be made in near real-time. Neither REST nor POSIX do this well, but is anyone working on this?
EMC helped start the world down the object path with the Centera product, but the open source world caught up to EMC. And it is very difficult to have a closed system software with just APIs compete with an open source software.
Though it looks like the FAA is going to delay UAVs flying in the USA until after 2015, that does not mean that the amount of video collected per day is not in the petabytes. And this will likely drive either a new interface or modification to an existing interface to better support the capture of that type of data.
I seriously doubt that POSIX will be modified so this leaves change to REST interface to better support video. Which brings me to my point: the newer vendors are seemingly ready and able to use open source technology, while the old guard is slower to catch on with using these technologies.
Maybe it’s because the new vendors do not have the development staff and cannot have a Not Invented Here (NIH) attitude, or maybe it is because they are closer to the market and can see the innovation coming. Either way the new guard has always been more innovative than the old guard. Very few of the new guard grow up and become the old guard. NetApp might be one of the few examples over the last decade or so.
Data Storage and Change
The old guard is not without tricks up its collective sleeves. In particular, buyouts. If you go back to the 1990s during the dotcom boom lots of the new guard companies got brought out and a number of them did not. Those companies that did not are most often not with us today. Many have gone bankrupt.
Historically, the old guard does not innovate, with few exceptions. But when the old guard does innovate it often falls short, given market perception. And NIH concerns about being locked into a single vendor solution surface, no matter how good that solution is.
The storage landscape is being pushed to change from two directions. The last few years we have heard about how flash is changing everything, and flash will replace spinning disk. This clearly will not be happening, yet there is a need for a high IOPS tier that is faster than 15K RPM 2.5 inch drives. What happens to software and hardware from the new guard now that Seagate has developed and is marketing an enterprise drive with flash right in the drive?
This Seagate technology enables the old guard to solve problems that they could not solve before, without making changes. The new guard sometimes does not have many of the enterprise features that make something really enterprise. Things like detailed SMART management and proactive error management T10 PI/DIF, replication, snapshots and other similar features.
Intel, with the move to putting the communications on the CPU, and Seagate with putting flash on the drive and building REST interface drives, both seem to be changing what the storage future might be. And seem to be trying to pressure the traditional players into adopting their new worldview.
In the 1990s, the old guard simply just purchased what they wanted and needed to be successful with Fibre channel, new file systems, and interfaces. The old guard is doing the same today. Yet the differences are significant and striking. They’re something that may or may not result in the same outcomes this time as the market dynamics are changing. Note to self: potential big changes ahead.
Photo courtesy of Shutterstock.