Samsung recently has discussed about so called “Storage Intelligence” at 2015 Flash Memory Summit which proved as a very interesting topic to globally located storage community, especially impressing SSD lovers.
The world renowned Korean brand said that it dreams of building processing power into modern storage devices such as SSDs and Hard disk drives which enable on-board encryption, sophisticated signal process and error correction. This gives us a feeling that the electronics giant is getting ready to design future storage devices in such a way that they can do more with the on-board processor in order to meet customer needs.
Traditionally speaking, in most microprocessors driven systems, which are slowly and steadily taking over the processor market, the main memory or digital storage is separate from the processor, connected to the processor by various communication channels. These communication channels are a cause of latency between a command and its execution.
Additionally, the host processor generally doesn’t know all the internal features of the storage device and so various translation layers and abstractions are necessary, creating even more sources of latency in working with the stored data.
This is where Samsung is gearing up to utilize the processing power built into a storage device, such as an SSD, to enable faster and more efficient applications by creating a tighter association between data and computational processing. The Tech giant calls this tighter association “Storage Intelligence”.
Speaking about Storage Intelligence and its features, Samsung said that the association between data and computational power offers better collaboration between SSD and storage system as it allows more advance garbage collection and better streaming of data by storing info in the stream in associated physical locations, which reduces overhead of garbage collection. It also allows offloading some operations, usually performed within the system microprocessor, to the individual storage devices.
For example the Solid State Drive can provide object storage by an internal key value storage API and by providing a framework for off-loading various types of processing to the storage devices. The on-board storage device processing must allow the host to control background operations. This avoids an overlap of processing and input/output (IO) operations that can degrade IO performance.
Thus, Samsung claims that its storage intelligence provides consistent, predictable and faster performance than the cases where computations are only done in the processor. It is because; the processing power gets accelerated by using unused processing capacity on the storage devices, and thus improves overall system performance and power efficiency, of microprocessors usually seen in mobile devices.
Subsequently, if the storage devices( like SSDs) are inducted with advanced processing power with microprocessors, then onboard object storage can be achieved on these IP based storage devices, as more advanced garbage management and higher endurance is gettable. This will offload processing from the main microprocessor to the storage device leading to more efficient operation and power savings.
Note- Already, IP storage drives are in the market in the form of Kinetic Drives. Seagate is offering these Kinetic Drives, which have an IP Address for them and can contact servers on their own, without the presence of any middleware server power.
Samsung wants to do more with this concept and wants to insert some more intelligence into this storage media, in a hope that it will do more by inducing onboard object storage into them in future.