Software defined storage (SDS) is a hot topic, a term that has been thrown around over the last couple of years and grabbed a sizeable share of the limelight.
As it grew in popularity, so did the myths that surround the concept. A recent study carried out by open source software company SUSE states that 70% of companies worldwide are concerned about the cost, performance and complexity of traditional storage. According to this study the scalability and efficiency of SDS interests a staggering 95% of these companies and nearly two thirds will adopt SDS next year.
Yet, there is still some confusion around what SDS really means. Why? Because some products are packaged as SDS when in reality they are not. Because it’s not always clear how SDS can make a difference to storage performance and capacity. And because it’s hard to separate the truth from the wild claims about SDS’s impact on the wider infrastructure.
“All SDS solutions are the same”
From a distance, it is true that SDS solutions can appear similar. In contrast to traditional storage where the hardware takes charge, SDS is simply a storage product where the intelligence lies in the software and is physically separate from the server and storage hardware. This decoupling yields cost, flexibility and scalability benefits.
That said, SDS architecture and implementation varies considerably from vendor to vendor. Some vendors provide only software and the user purchases the hardware, other vendors sell the SDS alongside preconfigured commodity hardware and others combine optimised hardware with their SDS solution.
“Any storage software is SDS”
It’s not unusual for solutions to be sold as SDS when they actually lack key features to make them such. Sometimes even the very basics, like the storage element, are missing.
Take, for example, products that employ software to create a single storage pool from several SANs, simplifying administration and increasing capacity. A more accurate term for this is storage virtualisation. Although these solutions are often marketed as SDS, the SANs provide the actual storage. True SDS brings the typical hardware functions, such as snapshots and replication, into the software layer.
>See also: Software is redefining IT infrastructure
It is hardly surprising that we see such variety in SDS solutions: there is still no standard definition. However, it is generally agreed that true SDS solutions have three common features.
First, storage services are abstracted and independent from the underlying hardware. Second, the platform is automated, simplifying management and reducing maintenance costs. And third, it has seamless scalability: scaling out as required without disrupting performance.
“SDS means I can have a unified storage solution”
While it is possible to combine block, file and object in one unified solution, it is not practical. Here’s why: imagine unified SDS as a Swiss army knife; it is a knife, a screwdriver, a pair of scissors and a bottle opener. It’s a brilliant concept when it’s in your pocket – however, none of these tools are as effective as the standalone versions.
The same applies to storage. Unified components don’t perform as well, so customers with complex needs should avoid combining block, file and object storage.
“Hardware does not matter with SDS”
It’s sometimes claimed that SDS will perform well whatever the set-up or hardware build. But, as counterintuitive as it may sound, SDS solutions are only as good as the hardware they run on.
What works for one user may not for another. As all SDS solutions are different, they vary significantly in terms of the hardware they require. Getting the right hardware can significantly reduce operating costs and provide a faster system with more predictable metrics, so it’s important to investigate before implementing new technologies.
“SDS does not work – it cannot provide the reliability of a traditional storage array”
This used to be true. Older SDS solutions were slow or unable to scale-out and were unable to compete with high-end storage arrays.
The next wave is, as the old business adage goes, “faster, better, cheaper” than high-end storage or all-flash arrays. Now SDS systems offer far more than traditional arrays in terms of reliability, performance, and scalability as well as cost.
The benefits of SDS are not just confined to storage. With software-defined technologies, compute, network and storage become tightly integrated. As a result, SDS is as much about computing and networks as it is about storage.
“Don’t I need dedupe and compression?”
Deduplication and compression are heavily advertised by all-flash array vendors. They are promoted as space-savers that reduce the data footprint, justifying the exorbitantly high cost of an all-flash array.
However, for the majority of high-performance, virtualised SAN environments today, adopting deduplication and compression results in sub-par performance at a higher cost.
Deduplication actually increases storage latency and requires significant compute resources. In terms of compression, there is not as much to compress as you may think.
Microsoft Word and Excel files, PDFs, photos – most of the files stored by an enterprise – are in fact already compressed. Instead, the SDS features that make a real difference in terms of performance are thin provisioning, snapshots and clones, and trim functionality.
Busting the myths
So, SDS is not a term to be taken at face value; it is often misunderstood. As adoption rates rise, it will be even more important to understand exactly what SDS can do and how to distinguish a real SDS solution from a technology that’s packaged as SDS but is far from it.
To do that, CIOs and IT managers must make sure they do their research when it comes to their next purchases. SDS offers far-reaching benefits in terms of cost, flexibility and scalability, however a solution must be chosen with care and attention or these benefits will not be realised.
Sourced by Boyan Ivanov, CEO at StorPool