Twenty seven years ago in 1989, I attended one of the very early virtual reality (VR) headset demonstrations in the UK. It was put on by a bunch of ex-INMOS engineers demonstrating the use of Transputers and Intel’s i860 to generate real time image rendering in VR environments, along with the first VR gloves.
Apart from the obvious VR wow factor, a significant memory of the event was someone falling off the stage as they lost their balance and orientation, which was quite impressive given the low resolution graphics at the time i.e. CGA, 640x200 pixels at 4-bit resolution. Luckily they were not seriously injured.
The killer app presented at the time was remote VR teleconferencing where individuals would magically appear across the table in front of you and be able to push an electronic document toward you which you could manipulate, read and mark up, all virtually of course. Wind forward to 2017. VR, thanks to dramatic advances in display technologies and smaller compact VR gear, is finally making it into some mainstream applications with far more realistic video and smoother graphics at a much lower cost point, along with a growing amount of web based or gaming content to fuel demand.
So why to do we care about this in the world of storage?
Facebook and others have for a while been claiming that VR will have a huge impact on data centers and telecommunications. According to Jay Paikh at Facebook: “By 2020, more than a zettabyte – that’s 1,000 exabytes – of information will be exchanged over telecom networks, much of it in data-intensive formats like video and virtual reality…” . Facebook are driving a new initiative to help address the network and communications challenges this will also present called the Telecom Infra Project. Intel, AMD and NVIDIA are also strong proponents of the technology given it drives the need for more powerful processors and graphics engines.
We are set for a very interesting transition for video storage over the next few years. VR isn’t a simple case of storing lots of HD, UHD (4K) or higher resolution video images anymore. We now need to store full 360 degree views for each video viewpoint to allow the viewer to literally turn their heads and look behind them while the video continues to run. This requires the video to be stored on storage that is highly responsive and able to shuttle the video smoothly in any direction. In case you’d missed it, the Google Data Center 360° Tour is a great example of how this works (remember to use your mouse to move the video perspective while watching if you don't have the recommended VR gear).
VR essentially means that we are going to need anywhere from 10-20x the storage capacity over a standard HD file depending on the video resolution, equipment and compression techniques used. For desktop gaming rigs, this maybe somewhat lighter as the various scenes can be rendered on the fly, but storage capacity and highly available storage (e.g. flash) are all going to make for a much smoother experience.
For real time 360 degree video like we see in the Google video , we are talking about a significant amount of real time sequentially streamed data that has to support rapid and unpredictable changes to the viewing pane. This requires access to fast, expensive storage. Yes, flash is getting cheaper, but as we’ve seen, flash storage systems are struggling to meet the current annual build out requirements of current storage let alone a 10-20x increase in capacity demand! This is one of the primary reasons hard disk technology will still be around for a while. The trick for data centers, along with individual gamers who are likely to be the early adopters of VR, is going to be how to balance the use of more expensive flash storage with much cheaper hard disk based technologies until the world has solved the cost, supply and demand problem associated with flash based storage.
Why we care about this at Enmotus is that a new class of storage virtualization and hybridization is required that can handle both streaming and randomized data. FuzeDrive™ can do just that. Just about all of the hybrid storage implementations out there today are based on some type of caching scheme which has been optimized around random, smaller chunks of transactional data and do not deal well with large streaming data files. FuzeDrive on the other hand, when paired with the right amount of flash or memory class storage, is able to address the streaming requirements of these newer, larger VR video streams as well as random data access requirements. Furthermore, it is able to address the balance of performance and cost automatically by ensuring that older, less accessed video and VR files are stored on less expensive hard disk or bulk flash storage.
If the predictions are correct around VR, data center storage requirements are in for some interesting challenges yet again over the coming years.
 The Virtual Reality Future: Bigger Pipes, More Data Centers by Rich Miller. http://datacenterfrontier.com/virtual-reality-data-center/