All-Flash Array: Looking at the Benefits and Disadvantages
With more storage and speed needed in the age of big data, all-flash arrays sound like the perfect solution to handle huge amounts of data. And since the prices for both flash NAND chips as well as SSDs have fallen over the years, these questions arise: What is an all-flash array? And why should I go for an all-flash array?
Simply put, an all-flash array is a solid state storage disk system that contains multiple flash drives instead of traditional spinning hard disks.
Well, the second question is not easy to answer! The best reply here is: It depends! That is because like everything in technology, it has it benefits and its disadvantages. All-flash arrays are targeted mostly for extremely high-transaction or throughput applications where speed and time are crucial. Financial trades, e-commerce sites, or on-demand streaming services, for example, would be examples of use-cases. Here the input and output (I/O) ratio the application requires is the main thing to consider and since the I/O rates of normal as well as enterprise hard disk drives remain unchanged and low for a couple of years now, all-flash arrays are the answer here.
Regarding I/O speeds per second, Dell EMC states that a system that needs 200,000 IOPS (inputs/outputs per second) to run an application on, only needs 10 high-end SSDs compared to 1,000 HDDs. Because of that, less servers are needed to have the same performance as before. Additionally, less energy is consumed by an all-flash solution and less storage administration time is needed to manage such a system. Time that, for example, in the case of a server farm, can add up to a lot of money being saved or spent otherwise during the year.
And with new designed storage controllers build inside modern systems, all flash arrays have all the necessary features that HDD based system nowadays have, too: deduplication, compression, and thin provisioning, as well as snapshots, clones, RAID, encryption, or API support. With built-in RAID support, the user is able to construct a normal RAID 5 out of several SSDs with the data security such a system provides.
So when I/O speed and time are the main benefits of all-flash arrays, what are the disadvantages, then?
That's quite simple: It's the price! Even though flash and SSDs have become cheaper, they are still more expensive compared to HDDs. So buying a new all-flash array can be a costly experience compared to a normal HDD system, which can only pay off when you use it for quite a long time.
The other disadvantage is that SSD based systems such as these all-flash arrays have a technology downside: Technology-wise are Flash chip best when data is only written on them once and then read many times. Flash chips only have a specified life span based on data writes. So the trick to make as few writing processes as possible to extend the life span of a all-flash array. The other technology problem is the SSD trim command.
The SSD trim command is a technique to avoid write inputs/outputs (I/O) that makes the life span of a flash chip shorter every time. Trim allows the operating system to flag blocks of data to be erased, when they have been released from the local file system. That means that the process of automatic erasing starts way in advance before new data is being written on the "empty" storage space.
Because of these facts - and to avoid loss of data that is needed in the future – business related or material that has to be archived because of compliance or legal reasons – flash based systems should only be used in environments were data is "only" read or analyzed. So using an all-flash array can be used and shows its enormous benefits in speed is for example a great idea for streaming servers like YouTube and others.
Also, using these machines for big data analysis or running huge data bases on them is a possible use case. For all data that needs to be either stored permanently or is being changed frequently, you should either choose a disk based storage, a server solution, or a hybrid storage solution which consists of both media types – SSDs and HDDs. Data that is analyzed or frequently addressed, but not changed, can run through the SSDs, while the other data is stored within the disks. Data that is most likely to be accessed is stored on the faster flash chips, while the other data is based on a traditional HDDs. This minimizes the chance that crucial data is lost.
However, in an event that data is lost in an all-flash array, there is still the chance that a professional data recovery service provider like Ontrack will be able to recover files. In many cases, the data structure is the same as with a normal low-end RAID systems and can be reconstructed. Additionally, the experts have special tools to rebuild a RAID-based or special high-end data structures, so that the data can be found and made accessible again. Therefore, in case of a data loss with an all-flash array, users should immediately stop running the system and shut down normally - then contact the specialists as fast as you can.
But as pointed out before, this is only possible when the built-in trim command in the system is either inactive or the space where lost data was originally located was not used for saving new data. Even then, the most experienced data recovery experts cannot do anything for you anymore.
Find more information on all-flash arrays here:
Picture copyright: Sebastian von Thadden / pixelio.de