2022 marks the 35th anniversary of the invention of NAND flash, a storage media that has transformed the way businesses consume and store data
According to Mordor Intelligence, the NAND flash memory market will reach a value of USD 94.24 billion by 2027, with a compound annual growth rate of 5.33% from 2022 to 2027. Despite this growth, the cost of solid-state drives (SSDs) remains a challenge for IT organisations and cloud service providers who need the performance that SSDs offer yet delay implementing due to the higher cost of acquisition.
To mitigate this, software solutions that intelligently manage SSDs can make a dramatic difference to the Total Cost of Ownership (TCO) of flash, and ultimately to an organisation’s margins. This makes implementing SSDs a much more palatable solution. This is possible by managing the writes to the SSD in a way that boosts endurance, performance, and reliability, enabling less expensive SSDs to reach the same levels as a more expensive flash device, while maximising the capacity available.
TLC and QLC
Since 1987, innovations in NAND flash led to denser and cheaper options, to meet industry demand to reduce costs and increase capacity. There are two main types of flash media available on the market today: triple-level cell (TLC) SSDs and quad-level cell (QLC) SSDs. TLC drives can store three bits of data per cell, while QLC can store four bits per cell. With the introduction of QLC, the price point of flash has dropped and made it possible to deploy flash at scale in disaggregated infrastructures. Yet the downside of this lower-cost media is that it is less reliable and more prone to errors, while endurance and performance suffer compared to more expensive TLC drives.
The life expectancy of an SSD is all about endurance, which can be optimised with flash management software to minimise the cost of regularly purchasing new storage media. Endurance refers to the number of programme/erase (P/E) cycles a flash cell can undergo before the storage media wears out and becomes unreliable. Each P/E cycle slightly degrades the oxide layer in the cell, causing leakage, and reducing the ability of the cell to accurately retain data. The way data is written to SSDs – sequentially or randomly, in large or small block sizes – will impact the endurance of a flash device. SSDs prefer to receive large sequential writes, while random or small writes will cause the storage media to wear out more quickly.
Application workloads often generate varying sizes of random writes, in different patterns. Even applications designed to write flash-friendly large sequential chunks are not ideal, as in combination these writes will be perceived as a random write workload by the flash device. In order to optimise SSDs’ endurance, software placed between the applications and the drives, organises the writes in a way that ensures the most optimal workload from applications, i.e. large sequential writes.
An additional benefit of writing data to SSDs in this way is performance. NAND SSDs have different performance values for sequential and random writes. Not only do sequential writes achieve 5x or 10x more bandwidth compared to random writes, but they also reduce latency. By boosting the performance of SSDs, the same hardware can serve a greater number of applications, minimising the number of SSDs an organisation requires and, ultimately, reducing TCO.
Additionally, intelligent flash management tools can ensure a consistent performance. No matter how strong average performance figures may be, if the performance is not consistent, it is inadequate to meet the requirements of modern applications. A few outlying slow I/O responses could negatively impact an online retail sale, a financial trade, or machine learning workload. The ability to guarantee tail latency is critical.
To further reduce TCO, it goes without saying that the capacity of the storage media must be utilised to the extent possible. As discussed above, intelligent software placed between the applications and flash media enables data to be written sequentially and in large chunks to optimise the SSD. The same goes for the size of the data sets written to the applications. Software that employs techniques such as compression and thin provisioning can more efficiently utilise SSD capacity. Flash management software solutions can compress data and aggregate multiple writes to create a single large write request that is sent to the SSD. This eliminates any wasted space.
In addition, solutions that offer thin provisioning can further improve capacity utilisation. Typically, a size must be specified when provisioning a data volume to an application, however it might take a while before that volume is completely filled up with data. Thin provisioning ensures that the physical capacity utilised from the system equates to the actual total data written, rather than what is provisioned. With a software layer placed between the SSD and the application, there is much greater control over where each 4KB block is placed, thereby enabling thin provisioning.
In large-scale deployments, hardware failures must be expected. And due to their composition, QLC drives are more likely to fail than other types of flash media. From day one, IT teams must select a solution that is designed to handle this with minimal disruption to the organisation.
Flash management software that offers self-healing can provide an important level of data protection should an SSD fail. It can ensure access to the data and keep the system running on other hardware, while repairing the SSD in the background. This fail-in-place concept avoids having to manually replace an SSD and can dramatically reduce potentially expensive downtime.
If you are looking to deploy QLC NAND flash, consider solutions with built-in flash intelligent flash management software. This type of solution can increase QLC endurance by up to 20 times. Look for a solution that is software-defined, which will provide the flexibility of hardware, simplifying supply chain challenges. And solutions that offer enterprise data services, such as data compression, elastic RAID, replication, snapshots, clones, quality of service, and multi-tenancy, will deliver greater storage efficiency, security, and availability.
The global datasphere is exploding. Organisations are generating and collecting vast volumes of data, but in order to extract the value from that data, they must access, analyse and store that data with limited IT budgets. By utilising flash management software organisations and cloud service providers can grow not only their top line by provisioning storage much more quickly to their applications and customers, but also improve their bottom line by maximising the capacity of lower-cost flash media and reducing downtime.
About the Author
Carol Platz is VP Marketing at Lightbits Labs. Lightbits makes high-performance elastic block storage simple, scalable, and cost-efficient for any cloud. We help our customers scale their business effortlessly, accelerate IT operations, and reduce costs.
Featured image: ©Vladimircaribb