How tape is opening up new frontiers
In fact, according to INSIC projections, by 2025 we may see HDDs of 40 TB in capacity, which is only about 20% CAGR in areal density from today’s products.
However ‘limitless’ the cloud may appear, therefore, any disk-based storage system will inevitably have to confront the reality that scaling pools of hard drives in object storage in the zettabyte era will need colossal quantities of devices, energy, floorspace and resource.
And none of those things come without strings attached. The world’s data centres already use more power than Spain. Already some newer data centres are being built with their very own solar energy source and I have even heard of studies exploring the feasibility of building underwater data centres powered by wave energy from above. So if it’s possible to reduce energy demands by using a cheaper and fundamentally more energy efficient solution, like tape, this has to be a consideration to evaluate against the perceived benefits of using disk.
And in terms of areal density, tape still has plenty of headroom. LTO-12 is targeting 480 TB per cartridge in roughly the same time frame of the next ten years. That’s around 33% CAGR in areal density - about double the projection for disk drives.
It’s important to realise that any disk-based array or object storage server using magnetic media technologies like HAMR or MAMR will be governed by the superparamagnetic limit. So for storing trillions of gigabytes of unstructured data in the very near future, tape is the ideal, cost effective long term solution. It has the support of leading vendors, it requires almost no energy at rest, it has a reliable roadmap and it’s already being used by millions of businesses.
Don’t get me wrong: the new HAMR and MAMR drives being developed by leading HDD vendors will be vitally important for the future of our digital universe. For the applications that require them - and there will be many - they will undoubtedly be deployed in colossal data warehouses like some of the gigantic installations being planned by cloud providers like Microsoft or Amazon. And those data centres will, over time, be powered by renewable energy sources which will mitigate the environmental impact of hyperscale storage and computing. For example, earlier this year Google announced it would install 1.6 million solar panels, equivalent to 300MW of power, to supply energy to two new data centres in the south eastern United States.
But if data is growing at 30-40% a year and disk capacity is only achieving 15% per year, then there is a fundamental mismatch between supply and demand. And the gap between those numbers is where I expect tape to come into its own as the most practical and cost effective solution for preserving infrequently accessed cold data for many years, even decades. The question is not ‘why would you consider using tape?’, but ‘why would you not consider use tape?’