Veröffentlicht April 24, 2023
Verfasser Jeniece Wnorowski
What if we told you that you are probably overpaying for data storage by as much as 20%? How is that possible, you may think. It can happen if the SSD you selected wasn’t exactly right-sized for your infrastructure.
Often, IT departments don’t fully understand how to choose the data storage best-suited for their arrays based on the attributes of their workloads. Similarly, by default, many storage architects design their infrastructure for worst-case scenarios.
But today’s workloads are changing, with most becoming increasingly read-intensive. In fact, data from Usenix, shows that most enterprise workloads break down to 94% read-intensive vs. just 6%  write-intensive. So, how should you choose data storage to meet today’s current workload demands? We’ll dive into this a bit later with a look at a real-world application. First, let’s see what’s changing in SSDs.
Organizations generally run a mix of workloads rather than any single workload type. Different applications require different profiles of performance, endurance, and, in some cases, different form factors. When organizations optimize for the applications they run, they often need to re-think and resize their storage needs. The days of “one-size-fits-all” SSDs are gone. Storage vendors who offer SSDs with a variety of performance and endurance profiles can help IT managers make better “total cost of ownership” (TCO) decisions, both in the short term and the long term.
If you are looking to preserve legacy infrastructure, U.2 capacity points have been extended for greater storage density. And, if you plan to fully modernize your storage infrastructure, the Enterprise & Data Center Form Factors (EDSFF) are becoming more widely adopted by the industry with each passing year. These newer form factors come in E1.S, E1.L, and E3.S variations. They range in capacities from 7.68 TB to 30.72 TB, with 30.72 TB as the highest-capacity PCIe QLC drive available today.
With an E1.L SSD, you can pack over three times more density across the front of a 1U rack vs. a U.2. You’ll also benefit from improved serviceability and greater space efficiency. By design, ESDFF provides better airflow, so it produces less heat than legacy form factors, which will also save you on energy costs, too.
Hyper-dense and affordable capacity Solidigm SSDs enable better performance and lower TCO benefits by reducing the number of drives and racks that servers need for the same amount of storage space. Not only that, but they require less power, cooling, and maintenance.
When SSDs were first deployed in data centers in the 2010s, the behavior of the drives—as well as the requirements of workloads—were not understood like they are today. With the benefit of hindsight, we now realize that storage architects tended to “over size” SSDs. This is obvious when we compare the average endurance level of SSDs shipped years ago to that of current models. Today about 85% of SSDs shipped to data centers have an endurance of ≥1 DWPD, which meets the needs of today’s applications at a lower cost than ≥3 DWPD. 
The reason for this right-sizing is that workloads and drive behaviors are better understood now. In the case of SSD endurance, studies with massive sample sizes show that 99% of systems use, at most, 15% of a drive’s usable endurance by its end-of-life.  Bottom line? QLC SSDs will not wear out as quickly as they once appeared to. Again, depending on your workload, you could be overpaying for data storage.
The good news is that affordable storage is also more sustainable storage. Let’s take a look at a large international streaming service, also known as a content delivery network (CDN). Its team members needed better performance and better capacity, and here’s how they benefited from implementing SSDs into their data server:
A top content delivery network provider moved from a hybrid storage solution of HDDs and TLC SSDs to an all-QLC mid-tier solution. It found that it could achieve a 4.9x server footprint reduction.  It used Solidigm QLC SSDs, which offered an alternative to HDDs and TLC NAND SSDs for affordably modernizing storage to better support on demand content delivery.
With read performance equivalent to TLC NAND SSDs, and much higher than HDDs, Solidigm QLC 3D NAND SSDs enabled storage architects to reduce TCO, efficiently scale content, and expand content to more users.
The combination of QLC’s smaller physical storage footprint and greater GB per pound/kilogram had huge implications for its data center design, becoming even more critical in its multi-story data center designs.
Higher density storage and a smaller server footprint helped to reduce total energy costs by as much as 4%.  This is a huge savings, depending on the size of the infrastructure, which this CDN was able to leverage when they decreased power usage and costs.
With greater density SSDs it can deploy fewer servers which results in a cooler environment. And, as we mentioned before, ESDFF form factors are designed for better air flow, amplifying that savings.
Higher density storage means 3.5x fewer drives to dispose of or otherwise disposition at end-of-life in that same mid-tier CDN solution. 
By deploying high-density, high performance data storage, you, too, can reduce your physical footprint and save on overall TCO. With Solidigm QLC SSDs, you no longer need to sacrifice cost for performance; or performance for capacity. QLC SSDs let you have it all. For more information on this topic, please watch the recent keynote at CloudFest, presented by Alexey Rogachkov, Solidigm Sales Director. Learn more about QLC SSDs from Solidigm
 Forward Insights Datacenter, May 2019. Approximate projected worldwide Data Center SSDs shipments 2020-2023 rated at ≤1 DWPD endurance.
 University of Toronto study of 1.4 million industry SSDs in Enterprise Storage Deployment. A Study of SSD Reliability in Large Scale Enterprise Storage Deployments, https://www.usenix.org/conference/fast20/presentation/maneas
 Solidigm. Solution requirements: A mid-tier CDN solution delivering BOTH 480TB of total capacity and 190 Gbps throughput per node. Source - Solidigm. https://www.intel.com/content/dam/www/central-libraries/us/en/documents/replace-legacy-storage-in-cdn-with-qlc-ssd-brief.pdf. See solution configuration details in Appendix A.