Until recently, the idea of using solid-state disk (SSD) flash drives in an enterprise storage subsystem would have been deemed ludicrous.
Ray Lucchesi, president of Silverton Consulting, however, says that recent trends in NAND technology have made SSDs more viable in the enterprise storage market.
In 2009 the flash SSD storage market will see even faster products appearing on the shelves, writes Samantha Sai for storage.biz-news.
That prediction is based on the additive factors that are at work in the storage market—architecture and semiconductor process technology.
Historically manufacturers did not see any point in integrating very fast architectural features in to flash SSDs as they were expensive and the market for the product was small.
The improvement in architectural technologies for speeding up performance and IOPS has since been developed by OEMs for specific marketable products and their potential is clearly exploitable in the Flash SSD design.
The SSD market has also reached a critical mass.
Lucchesi said the technology has now become economical enough to favorably compare to traditional disk drives — at least in a price-performance context.
In addition, he said many seemingly insurmountable shortcomings have been resolved.
The architectural features that would come into play in the construction of Flash SSDs would be:
- parallelization of the internal media arrays
- improved media management technology
- faster media controllers
- faster host interface controllers
- hypbridizing on board memory technologies
The scale up of the technologies would require significant investment in IP. A lot of trial and error will become visible as OEMs throw products at the market.
In this context it is interesting to note that it is expected that a single 3.5" form factor flash SSD will be able to deliver speeds of 2000 MBs of sustainable reads and writes.
The Flash SSD through put and IOPS performance is predicted to be a multiple of the performance for a single disk, based on the proven scalability of the SSD RAID arrays.
Expectation is also high that the asymmetry of sustained read to write IOPs will improve from 10 to 1. (However, there is apprehension that it will never achieve parity.)
If this expectation is fulfilled, flash SSD arrays will become a viable choice for many Enterprise speed-up applications.
While the latency in Flash SSDs is not expected to scale in the same way as throughput, the read access times are expected to improve.
This is attributed to the fact that flash SSDs have not yet been optimized for latency. However, in the future the Flash SSD may increase in density and a read write cycle may become more complicated on chip process.
Calibration, error correction and address translation may be done by controllers between the memory arrays and the host interface controller or card data bus.
The flash may still evolve as a separate species that looks completely different from the typical RAM.
On the whole, the improvement will be revolutionary and evolutionary and by 2013 flash SSD would have reached a point where architecture of an ideal SSD will be well established.
Comments