SSDs And Steve Wozniak: A Longstanding Appreciation For The Clever Hack

Speaking of Fusion-io, some of you may not already be aware that Steve Wozniak is now the company’s Chief Scientist. Wozniak aka ‘Woz’ is, of course, co-founder of Apple and the company’s first engineer…although younger readers may remember him more for his sponsorship of the US Festivals in 1982 and 1983:

or perhaps his more recent stint on Dancing With The Stars:

When I spoke with Micron on Tuesday, I suggested to VP Dean Klein that the company might need to one-up Fusion-io from a celebrity staffing standpoint…might Steve Jobs be available? And when I met with Fusion-io a few weeks back, I asked company officials what was at the core of Woz’s interest in them and their products. The story they briefly shared with me is reminiscent of the hardware-versus-software SSD architecture differentiation between Micron and Fusion-io that I talked about in my previous post. And I’ve subsequently been able to confirm its validity thanks to an excerpt from Wozniak’s ‘iWoz’ autobiography, posted on Amazon’s website.

The tale, in brief, involves Wozniak’s request to accompany Steve Jobs and Mike Markkula to the 1978 Consumer Electronics Show, held less than two years after the company’s incorporation. Woz was told he could go if he figured out how to mate the just-introduced Apple II computer with a floppy drive. CES was in less than two weeks, and Woz was chartered with designing an add-in controller card for the just-introduced Shugart 5-inch format drives. I’ll let you read the full fascinating excerpt, which spans less than 10 pages and includes an inadvertent floppy disk wipe of the source code, on Amazon’s website. But to whet your appetite, here’s a brief quote:

In the end, I decided that of the twenty-two or so chips, about twenty of them weren’t needed. To make the floppy disk work required a combination of a circuit I had to design and the existing circuit on the Shugart drive. I stripped out twenty of their chips, so that reduced twenty of my total end product. That’s the way I always think about things. I could run data right from my own floppy controller to the read/write head and implement any start/stop protocols of my own in code on the computer. To tell you the truth, it was less work on the computer than generating the funny protocol Shugart wanted.

In pondering the divergent Micron-vs-Fusion-io approaches earlier today, I drew an analogy to PC audio and its evolution over the years. Not too long ago, in order to achieve a robust audio presentation coming out of your computer (virtual or physical surround sound, high sample rates and large sample sizes, minimal ‘lip sync’ delay, etc) required a beefy standalone DSP. Commensurate with the publication of Intel’s AC’97 standard, however, silicon’s grip on sound began to slip.

The analog-heavy portions of the audio subsystem split out into a separate codec IC and, beginning with low-end systems, the digital audio processing algorithms ran on the host CPU instead. And as x86 CPUs became increasingly robust, processing delays actually decreased with a software-centric solution, versus the traditional approach of shoving a bunch of code down a slow peripheral bus for subsequent audio co-processor attention. Very few audio add-in cards survived the PCI to PCI Express transition, and nowadays host-based audio processing reigns supreme, lingering dissenters be damned.

RAID provides another example of the software-vs-hardware tug-of-war. While it’s still relatively common to find hardware-accelerated RAID add-in cards, especially for more complex RAID levels that encompass greater than two HDDs or SSDs, two-drive RAID 0 and 1 are commonly implemented in software either at the BIOS level or within the operating system. Arguably, assuming your system is sufficiently CPU-robust to handle mass storage subsystem administration without discernibly slowing down in other areas, software-based RAID (especially at the O/S level) is the preferable approach. If a hardware RAID controller or motherboard dies and you can’t find a replacement, you’ll never be able to access your drive-based data again. Conversely, with a standardized O/S-based software RAID approach, you never have to worry about hardware failures locking you out from your valuable bits.

So which approach to flash memory array management, hardware or software, will reign supreme both in the near-term and over time? Should be fun to watch. Speaking of watching, I’ll close out this post by wishing Steve Wozniak a belated 60^th happy birthday; it was Wednesday. You advance the profession of engineering for us all by your enduring example, Woz. Thanks to a Cult Of Mac posting, I’m able to share with my readers a birthday song commissioned by Woz’s wife and performed by talented and prolific musician Jonathan Mann, whose ‘Antennagate‘ video I showcased a few weeks ago:

Enjoy, and have a happy weekend.