TORONTO – A startup has an idea for a persistent memory controller it believes will help get the most of performance from a storage class memory (SCM) by doing away with translation tables.

Wolley’s patented table-less architecture was developed with SCMs in mind, said president and co-founder Bernard Shung, in a telephone interview with EE Times. It offers the potential of faster performance by using fewer states than existing controllers by removing the need for a flash translation layer (FTL) address table. The technology also eliminates the need for a large backup energy cap, he said, and through its design, fundamentally resolves the difficult issue of synchronization.

Shung said that Wolley’s technology reflects the company’s conviction that an SCM requires a fundamentally different controller, one that is table-less. With speeds approaching that of DRAM and the capacity and non-volatility of NAND, an SCM has been a “dream device” for decades. Added Shung, “It wasn’t until Intel/Micron announced 3D Xpoint in 2015 that people started to pay serious attention to it.”

Although a typical SCM has better write endurance than NAND, it still needs a controller to perform address translation. A SCM device performs wear leveling by mapping the host address into another physical address. Traditional NAND controllers employ an FTL, which uses an address translation table for the mapping between host address and physical address. The table access is an extra step, said Shung, and the table can either be stored in DRAM, which is faster and more expensive, or in NAND, which is slower and cheaper, or a combination of both.

A notable cost factor for storing the table in DRAM is that during a power outage, the DRAM content needs to be saved in NAND, often requiring a Super-cap to hold the power temporarily, said Shung. The Super-cap could be expensive and sometimes bulky if the DRAM table is sizable. “Eliminating the table gets rid of the DRAM and Supercap, if the table is to be stored externally. It also eliminates the extra SCM access step, if the table to be stored inside the SCM, and hence the performance is better.”

Because of these challenges, Wolley began exploring a table-less SCM controller architecture a couple of years ago. “We believe a table-based FTL is a good choice for NAND controller, but a table-less controller architecture is better for SCM.”

Shung said for the immediate future, Wolley is approaching all SCMs as if they were similar, but capacity and write endurance will be distinguishing characteristics as the company looks to productize its controller.  

Acting AG Whitaker

Wolley is not the first company to come up with the notion of a controller without a translation table, noted Jim Handy, principal analyst with Objective Analysis. About a decade ago, IBM Research came up with what it called “Start-Gap,” a way of moving data around without needing a translation table, but by using different algorithms. IBM actually came with up with term “storage class memory.” Added Handy, “They have been very focused on the idea that storage-class memory is coming for a very long time.”

Having a controller without a translation table is a good idea, said Handy, because the “big deal” about an SCM is it’s fast enough that you don't want to have a translation table get in the way. “Whenever you bulk up the speed of any storage-class memory — MRAM, XPoint, or PCM, anything like that — then all of a sudden, the speed of the translation table becomes a big headache. This gets past that.” Handy said such a controller would probably work with any SCM.

When SSDs first came on the scene, people putting them in systems found they didn’t get the anticipated speed improvements because the software used to communicate was slow, added Handy. Tuning was necessary so software stopped being a bottleneck that limited the performance benefits of NAND, and a controller without a translation table has the performance improvement potential for SCMs.

Wolley’s controller technology is still at the prototype stage, but Handy said it’s conceivable that a larger player could scoop it up, as Apple did with Anobit and SK Hynix did with Link A Media.