With SSDs, Multi-Level Cell Designs Might Be the Way to Go


There are all sorts of questions surrounding the deployment of solid-state disks (SSDs) in enterprise settings. But none is more fundamental than the basic dilemma of which kind of SSD to invest in.

Do you go with the higher-priced single-level cell (SLC) to maximize the performance gains the technology brings over conventional magnetic media? Or do you try out the new lower-cost multi-level cell (MLC) designs that offer more capacity at the expense of throughput and durability?


The answer to that question is as varied as the number of drives on the market right now. But a growing chorus of opinion is starting to center on the notion that, despite coming at a premium price that is not likely to hit comparable levels any time soon, the advantages of SLC may not be as clear-cut as once thought.

Objective Analysis' Jim Handy has a good run-down on the advantages of MLC this week on Infostor. His argument is that performance will be adequate for the majority of enterprise needs, and that advances elsewhere in the storage architecture can make up for any deficiencies. He writes that the latest controllers are perfectly capable of compensating for both the endurance and I/O speed handicaps that come with MLC. And the speed factor in particular is of minimal concern because it generally only affects writes, so it makes sense to deploy MLC more broadly considering it costs less than a quarter of SLC devices.

Because MLC can be designed in many different ways, the type of configuration on the die can have a substantial impact on performance, particularly as manufacturers shrink the process down. Intel and Micron, for example, have chosen to go with two bits per cell in the new 25 nm device, which they say provides an even split between the speed benefits of an SLC design and the capacity advantages of a 3-bit MLC approach. The 2-bit solution delivers 200 MBps using the ONF1 2.2 source synchronous interface and can provide 256 GB using only 32 dies.

The newest generations of MLCs might even take the endurance issue off the table entirely. Smart Modular Technologies made headlines earlier this year with the new XceedIOPS drive, which not only pushes capacity up to 400 GB, but improves the longevity of MLC operation from the current 5,000 P/E cycles to more than 20,000 P/E cycles. The unit features the new SandForce SF-1500 processor that delivers 250 MBps sustained throughput.

The same processor also powers the new TeraDrive FT2 from Super Talent, which operates on a 3 Gbps SATA interface and features the company's DuraWrite technology that extends drive lifecycles through a combination of wear-leveling, block management and advanced error correction. The company claims an MTBF of more than a million hours.

In all likelihood, there will be specialized circumstances in which SLC performance is called for. But it looks like that for the majority of enterprise operations, MLC will fit the new mantra that's taking over: not optimal, but good enough.