Toshiba Electronics Europe gives a prediction: what the storage trends will be in 2019

Where are we going to store so much data? Thanks to engineers and programmers, the capacity of disk media is getting larger, and combining it into efficient storage systems is getting easier. So the problems we face today are easier to solve. But with predictions of rapidly accelerating volumes of data generation by machines such as autonomous vehicles and smart factories, and given that humans have already created massive amounts of data, including backups, will we be able to produce enough storage for the needs of the next decade? Or we’re going to have to think about a more ruthless approach and decide what not to store?

Author: Reiner Kese, senior manager, storage business development, Toshiba Electronics Europe

The balance of HDDs and SSDs in a world of growing data

Not only is the amount of data stored in the world still growing, but it is growing faster than predicted. The proportion of data stored on flash and SSD was expected to increase, while the amount of data stored on hard disks and magnetic tape was expected to decrease. Today, however, it is clear that all three technologies continue to evolve simply because so much storage capacity is required. We can assume that in 2019, 90% of the capacity for typical cloud computing applications will be provided by hard drives, perhaps partially on magnetic tape. And only 10% will be provided by SSDs. But because

enterprise SSDs cost up to ten times more per unit capacity than HDDs, the financial investment will even out: about 50% will be spent on HDDs and the same amount on SSDs. Storage systems like these span the entire spectrum of applications. Some will use flash exclusively, some will be hybrid models where flash stores cache or “hot” data and hard drives store “cold” and “warm” data, the rest will be servers based solely on hard drives.

Helium HDD capacity to grow to ~20TB

All three major manufacturers have now introduced helium-filled hard drive models with capacities up to 16 TB. Capacity is expected to increase by about 2TB per year over the next few years, meaning that 20TB hard drives should be available early in the next decade. These drives are likely to lose in value as capacity increases, but other technical parameters are not expected to improve markedly.

The only exception will be energy consumption, which will be reduced by the use of helium. While air-filled 3.5-inch hard drives running at 7200 RPM consume, regardless of capacity, a relatively constant 11 watts under load, the power consumption of helium-filled hard drives is about 6-7 watts. This is a result of the lower friction created by helium’s lightness. Thus, the proliferation of helium hard drives will help solve the problem of growing data center power consumption.

Every watt of power saved by these drives reduces the amount of power needed to run the data center and also reduces the heat dissipated, making cooling more economical. Another effect of lower temperatures: helium actuators have higher reliability than air-filled actuators in continuous operation. This results in far fewer failures and a longer lifespan. Further storage densities are also planned, thanks to technologies such as microwave assisted magnetic recording MAMR , which will be integrated into the hard drive’s write heads.

Storage architecture

We can expect to see continued growth in the top-loading rack storage sector because of the demands for increased capacity. Although 60 bays in a 4U format has now become the standard, cases supporting 78 to 110 bays for 3.5-inch hard drives have emerged. Instead of hardware RAID, so many disks are configured with software-defined solutions. Modern software-defined storage will continue to dominate, along with scalable projects such as Ceph clusters, where multiple storage servers are combined into larger blocks. Here data protection is no longer ensured by excessive hard drives on the server. Instead, redundancy is realized through storage server nodes available on the server network.

Explosive growth in the amount of data

Already today, humans are generating huge amounts of data. When you consider that this data is then replicated in processing centers and in the cloud, it only increases the amount of memory needed to store it. The amount of machine-generated data today is relatively small. However, this will change starting in 2019 as solutions and technologies such as autonomous cars, smart factories, the Internet of Things IoT and home automation will generate additional

Flows of data that will have to be stored. Their expected volume is so large that the current storage philosophy needs to be seriously reconsidered. The harsh reality is that we need to analyze the data before saving it to determine which ones are really important and cannot be deleted.

Artificial intelligence, deep learning and blockchain

New computing applications such as artificial intelligence, deep learning and blockchain have greatly increased the performance requirements of data processing systems. We can expect these technologies to generate a lot more data and need storage solutions. At this time, it is unclear exactly what impact they will have on repository requirements, as not enough is known about the applications themselves and how they will be implemented. However, as we approach 2019 and the new decade, the situation will become clearer. It is already clear that these technologies will further increase the amount of stored data.