Software-Defined Storage (SDS) is a frequently used term that can have different definitions depending on the vendor using the term. SDS solutions can be block, file, or object-based. File systems are almost always software-defined and nearly always run on top of an operating system. The implementation method that file system vendors choose impacts the overall performance of the solution. The choice also impacts the vendor’s ability to support new hardware as it comes to the market and the options the customer has when it comes to hardware selection.
File Systems and Operating Systems
Typically, file systems run on top of an operating system, the most common of which is Linux. Linux supports a wide variety of servers, storage, and network hardware. File systems that run on top of it inherit that compatibility. Also, when new hardware, such as next-generation CPUs, Input/Output (IO) buses, or storage media, come to market, the file system vendor can almost instantly support these advances.
How open the file system vendor makes their software to the available hardware varies by file system vendor. Some vendors provide the customer with no choice, and force them to buy a single turnkey (software and hardware) solution. The turnkey approach leads to high-quality support and high levels of compatibility but does limit customer flexibility. It may also increase costs.
At the other end of the spectrum is a software-only solution that enables the customer to run the file system on any hardware they choose. The problem with the software-only approach is hardware qualification and selection is entirely on the customer. IT may not have the time or experience to qualify hardware properly, and a bad selection may lead to disappointing results.
A more balanced approach is for the file system vendors to use a reference platform approach. With this method, the file system vendor qualifies a specific hardware configuration available from several competing vendors. They may also qualify their solution to run in several of the cloud providers. The result is the customer maintains their freedom of selection while the vendor can also assure high-quality support and service.
Vendors may also choose not to use a conventional operating system such as Linux and instead develop a proprietary one that is focused more on IO performance. The proprietary operating system’s IO focus does seem to deliver more IO performance than many of the Linux-based competitors. To combat the potential IO bottlenecks that Linux introduces, a few vendors are bypassing the Linux IO stack and directly managing, reading, and writing to the storage media. The direct communication with storage media gives these Linux-based file system provides the best of both worlds: the broad compatibility of Linux without having to compromise on performance.
The increasing use of artificial intelligence and rich media is making organizations demand more from unstructured data storage infrastructure than it ever has before. IT needs an infrastructure that can deliver high performance and massive scalability, while at the same time reducing the cost to retain this data. Vendors that use Linux as their base, enable their file system to support the broadest range of hardware, allowing them to create solutions that meet these demands. It is critical, though, that the flexibility that Linux brings doesn’t limit performance.
Watch the below Lightboard Video as George Crump, Lead Analyst at Storage Switzerland, and Jason Sturgeon, Senior Product Manager at Qumulo, discuss how to use a balanced approach to enable a file system to meet all of its customer’s challenges.