Part3 of the #thinksustainable series
According to Marketsandmarkets™, Hyperscale data center data storage is growing at over 16% CAGR1. Unlike compute that has elasticity that grows and shrinks as the performance is required, storage is persistent. Persistence of storage not only spans capacity, but it also spans durability, and availability. It must be durable across data stores, zones and geographies. In short, the amount of data stored will always grow, and for service providers and data center managers to monetize data, retention must be managed in the most efficient manner.
The first measurement that storage admins think about is Total Cost of Ownership (TCO), the cost per unit to store the data over the long term. The TCO for storage media has been well defined and often overused to establish priorities for storage. Overused simply because it is well established that archive data is the lowest cost storage available today and in the foreseeable future. That leads to the fundamental question of what else can be used to measure the efficiency of archive storage, especially as tape is 4 to 8 times less expensive than HDD.
What about the density of storage?
As with compute resources, the density of storage is critical in Hyperscale environments. Maturity in “All HDD” managed service providers has led many to not only consider but deploy tape. This started as a pure TCO play but has quickly developed into a story of the density of storage for archive data. While data storage is growing at 16% CAGR the archive component of that data is estimated to be growing at up to 35% CAGR1. For MSPs to be successful they must embrace new methods of storing archive data.
In a 2020 blog2, Backblaze announced that their data storage service had reached 1 Exabyte (EB) under management. Backblaze is using an industry average of 14TB HDDs to store that data, 122,658 HDDs to be specific. In the OCP Bryce Canyon design, HDD density is at 72 units per drawer with up to 10 drawers (including controllers) in a standard OCP rack. The most important number is that his amounts to 171 racks of HDD storage. In part 2 of this series, we discussed the energy consumption of the OCP Bryce Canyon, now we will compare this exabyte to a tape storage solution.
The IBM TS4500 HD solution deployed as an available data archive, utilizing LTO-9 technology can store up to 22.5 PB per OCP rack equivalent. As a result of this extreme density the IBM TS4500 solutions can store 1 EB of raw data in only 57 OCP rack equivalents.
The future is now
The high density of the current IBM tape technology is just the beginning to establishing hyperscale archives. IBM is leveraging today’s technology to introduce the Ultra Density (UD) tape automation, capable of storing 1 EB in 38 OCP rack equivalents, and at the same time boosting access performance by 1100%, using LTO-9 storage capacity. At this density, IBM will be enabling just over 4.2 EB of data center configured archive storage with over 3.2 PB/hour of data throughput in the floor space of the average American home.
- Marketsandmarkets™ - Next-Generation data storage market with COVID-19 impact analysis 2021
- https://www.backblaze.com/blog/what-is-an-exabyte/