More corporations than ever are taking advantage of flash storage, and you use it every day – be it in your computer, phone, wearable device, or in the USB drive you back up your data on. It’s faster, more durable, and more reliable than a spinning disk, but it still does have one weakness: it deteriorates over time.
In theory, all of the flash’s advantages eventually go out the window as its lifespan wanes, and since it’s more expensive than a spinning disk, experts are trying to find new ways to make it live longer.
How flash storage deteriorates
To put it simply, flash storage is made up of a group of cells. Each of these cells are individually written with data, and then hold this data for you to reference later. Your pictures, documents, and videos are all broken up into several tiny pieces and individually written into these cells, similar to writing something on paper with a pencil.
Like pencil and paper, when you erase what the cell held – removing a photo or document – and replace it with something else – a new photo or document – you’re forcing the cell to rewrite itself, like erasing pencil lead from paper. If you use the eraser too much, the paper will thin out and eventually rip. That is the Achilles Heel of flash storage.
How experts are making flash last longer
1. Machine learning adjusting the load to make flash cells last longer: There are high points and low points in everything’s lifespan, and flash is no different. The kind of load it can take early in its life is going to be different from later in life, with a few ups and downs in the middle as well.
Machine learning is being developed and tested by experts to automatically identify when these high and low points are, and purposefully lighten the load or distribute the strain rewriting puts on flash storage cells so that they’re not overly taxed, which wears them out sooner. This process is so specific, so fast, and so difficult to detect that it’s impossible for human beings to do it manually – which is exactly what the computer itself would be set to the task.
2. Removing static data from wear levelling flash storage: There is static data and then active data, both which live up to their names.
Static data, when written on a cell, only needs to be written once – so those cells can take a break; their job is done. However, active data, such as a program is being written and rewritten often, makes those cells work constantly. To keep any one cell from being worn out too fast, wear levelling is a built-in flash feature that spreads that weight out over multiple cells.
The problem is that static data is makes it impossible to spread that wear evenly – it’s not active, so it can’t be moved. This causes the active data cells to wear out considerably faster, and when they wear out, the entire thing wears out, dramatically reducing the lifespan of the drive. That’s why experts have created wear levelling and non-wear levelling flash storage, so active data and static data can each have their own storage to avoid overburdening the flash.
3. Digital signal processing to make bit errors more readable: When the cells are written and rewritten too often, they can misunderstand the finer details of what they’re supposed to write, which essentially makes them mistranslate the data. This is called a bit error, and the cells have to work harder to decipher what the data really means, putting a strain on them and their lifespan.
However, experts have developed storage systems that use digital signal processing (DSP) that takes half that burden on itself, splitting up the strain so that bit errors can be avoided or handled faster, putting less strain on the flash storage itself. However, this is a temporary solution that extends flash life only to a certain degree.
Conclusion
Flash storage is like all things; it’s mortal. It’s impossible, for now, to create flash which doesn’t wear out, but in the meantime, experts are tailoring certain storage to help spread out the work and bear the additional weight, and even teaching the flash storage itself how to better allocate its efforts in converged infrastructure to keep it fresh. The longer the storage lives, the more cost effective it is, which inherently drives more funds to improving its lifespan.
[Source:- cloudcomputing-news]
