Brian's Brain

Back on May Day, I discussed (among other things) the fiscal foundations of flash memory manufacturers' robust motivations to ensure SSDs' success. In mid-June, I followed up with a specific example of the technology's worrisome over-supply potential; a 16 Gcell (32 Gbit) MLC (multi-level cell) NAND flash device fabricated on a 34 nm lithography process co-developed by Intel and partner Micron Technology. And today we get the next iteration of Micron-branded storage subsystem output from the partnership (albeit not based on the 34 nm memory, since it's not yet ready for production), as Micron unveils SLC (single-level cell) and MLC SSD variants in both 1.8" and 2.5" form factors.

My briefing with Micron last week was staffed by (among other folks) Micron's Dean Klein, the company's Vice President of Memory System Development and someone whom I've been working with since the days of the DDR SDRAM-vs-Direct Rambus DRAM wars. Dean's on my short list of highly desirable spokespersons who are both:

• Technically adept (for example, his Rendition-rich team developed both core logic chipsets and motherboard reference designs derived from them in order to seed the PC OEM and ODM markets with DDR SDRAM-based technology), and
• Straight shooters

although not everything he shares with me is 'on the record' ;-) As such, I peppered him with quite a few questions not only on Micron's products, but also on the big-picture SSD-vs-HDD market.

The company's first-generation SLC-centric SSDs, the C100 series, were based on a Micron-developed memory controller. Lingering SATA PHY and other circuit shortcomings, in spite of multiple silicon revisions, has led to their early obsolescence in favor of the C200 (MLC, note the technology change within the 'C' family moniker) and P200 (SLC) successors. These second-generation RealSSD offerings (with Q4 availability but, unfortunately, with no published pricing) employ an un-named partner's memory controller, albeit one running Micron-developed software, supporting 3 Mbps SATA transfer rates. The physical design of the P200 and C200 drives is identical; performance and cycling endurance differences between them reflect their respective SLC-vs-MLC flash memory technology underpinnings along with firmware variations between them.

The P200 series' key target is the enterprise. The SSDs' transfer rate performance versus 10,000- and 15,000 RPM HDD competitors certainly stands out, as does their nearly instantaneous 'seek time' versus the rotating platter- and moving head-based HDD alternative, the latter particularly notable when multiple network clients are simultaneously contending for storage access. And IT will welcome SSDs' comparatively low power consumption (i.e. heat dissipation) under constant-access enterprise conditions, as well.

Speaking of HDDs, Western Digital's latest VelociRaptors bear consideration in enterprise applications, too. They're passive cooling-less versions of the consumer-targeted drives I talked about back in June, therefore in 2.5" form factors versus 3.5" for the earlier-mentioned products, and they also tout beefed-up MTBF specifications. Check out HotHardware's study to see how three VelociRaptors perform in a RAID 5 configuration. And if you're curious to see how WD's speed demon stacks up against a high-speed SSD alternative, HotHardware can help you here, too. VelociRaptors deliver higher absolute capacity and (presumably) lower cost-per-GByte metrics than Micron's SSDs...although on that note, it should be mentioned that to minimize HDDs' power consumption and latency delays, enterprise applications sometimes use only a fraction of the available rotating magnetic storage potential, focusing on the outer-rim portion of each HDD platter.

One key enterprise feature that both Micron and WD's products lack, however, is SAS capability. As I wrote back in early 2004, SCSI command set support is deeply embedded within many IT applications and is therefore a no-negotiation requirement for consideration of any new storage offering. Micron's Klein was non-committal when I asked him about the company's SAS plans; if Micron didn't offer SCSI-friendly products, he noted, a partner company using Micron-fabricated silicon could alternatively supply this particular market segment. Klein also noted that SAS-to-SATA bridge ICs were also available; this is, for example, how many first-generation SAS HDDs (using a SATA HDD foundation, reflective of much higher SATA demand) were implemented. If NAND flash memory vendors' dreams are answered and supply goes constrained, though, I'm not sure if I'd want to be in a partner company's shoes...Justin Sykes, Micron's SSD Director of Marketing who was also on last week's teleconference, was adamant that in such a case, 'we'll supply ourselves first'.