Flash storage and the 80/20 rule

Flash memory based storage is starting to make serious headway into the PC realm, and with good reason.

It’s far more reliable.  It can be squeezed into many different form factors…from 2.5″ notebook drives all the way down to MicroSD cards.  It consumes almost no energy and is completely silent.  The random access times obliterate traditional hard drives and, while not quite on par yet, raw transfer rates are now beginning to close in.

To dispel a common myth, one thing you commonly hear about flash memory is how it’s limited to around a million write cycles.  So eventually, the memory will die.  Yes, this is true.  However, objectively speaking, when you look at the typical lifecycle of a hard drive, a million write cycles actually buys you much more time than the MTBF (Mean Time Between Failure) of a hard drive.  Flash memory is just being more up front about it.

Flash drives use all sorts of tricks and have some extra capacity in them to distribute the write load across different blocks.  This is called wear-leveling.  Furthermore, the roughly million write cycles required to wear out a block is a gradual failure.  What you’ll usually see happening to a flash memory drive is a gradual loss of capacity as impending block failures are detected and rerouted.  Traditional hard drives are not nearly as nice about failures.  They tend to blow up all your data at once…which, in my humble opinion, is a big nail in the coffin of PC friendliness.

Now, the upsides of flash memory are obvious, but so are the downsides.

As I mentioned, the raw transfer rates are in the same order of magnitude, but in most cases flash drives are currently a good deal slower than mechanical drives.  The biggest problem, by far, is cost.  Flash memory still costs far more than magnetic storage on a dollars per GB basis.  The first 64GB flash based hard drives to be released will probably cost around 2000 dollars.  Meanwhile, you can buy a traditional 500GB drive for just 180 dollars.  Hardly seems appealing in that context!

Now, both flash memory and magnetic storage technologies continue to improve.  But it’s interesting to note that if you look at a 12 month magnetic storage cycle, flash memory is improving on an equivalent basis in just 9 months.  Obviously, this means at some point, flash memory will overtake magnetic storage on a pure cost per GB basis if the cycles continue to hold.  But we’re still a good way off from that happening.

Still, aside from the cost per GB, the benefits of flash memory may show up sooner than you think.  Instinctually, you might think…2000 dollars for 64GB?  What the heck?  Sure, the tremendous amount of storage we get from magnetic drives is not something people will give up easily.  But stop and think about this for a second.

1. Only a relatively small portion of data needs to be cached to avoid hitting the mechanical drive most of the time.  This is where the 80/20 rule comes in.  In general, e-mail, source code, documents, and other work related data simply don’t take a lot of space.  I’m guessing that the OS + most applications + application data can easily fit into 16GB for most people.  Maybe bump it up to around 64GB to be safe.

For large files like audio and video files…well sure, keep them on magnetic drives and just spin those up when you use them.  For most people, this won’t be very often.  For all the times in between, several GB of  flash memory augmenting a system will draw less power, run apps and boot faster, and keep running more reliably.

2. Flash memory doesn’t have to replace mechanical storage.  Because flash memory is easy to produce in small form factors, it’s going to be very easy to add flash memory as a complement to any desktop or portable system.  A good system will play to the strengths of both.  Simply use as much flash memory as you need to get a nice performance boost, but not more.  And continue using mechanical drives to store large quantities of data in a cost effective manner.

And at this point, you’re not really talking about dollars per GB at all.  You’re asking a different question…how much is someone willing to pay so that

1. The computer loads programs more quickly and boots much faster
2. The computer is less noisy.
3. The computer, if it’s a laptop, has more battery life.

An entire community of users exists that is quite willing to pay a lot for any possible performance boost.  Using flash memory in a system unlocks new performance characteristics that have previously been unavailable to users.  Some of these users are going to bite.

Dollars per GB is important as a metric for flash memory in some respects, of course. But given the above discussion, you can imagine the following use cases for flash memory.

1. Performance/power/silence booster for systems
2. Convenient way to add storage (possible, not sure if it’s likely)
3. Mass storage replacement for mechanical drives.

1 will obviously happen much earlier than 3 due to cost considerations.

In almost all cases right now, the cost of flash memory is currently prohibitive. However, I think that if you figure out what amount of GB in flash produces the bulk of the performance benefit, and then you figure out what that performance benefit is worth to your target market, you start to get a pretty good idea of when flash memory is going to hit its first breakpoint in the PC market.  The second breakpoint will happen when flash memory starts to reach cost parity with mechanical storage + some added premium for the performance advantages afforded by flash memory.

Enough expounding on the subject for today.  I hope some of you found that interesting.

This entry was posted in Uncategorized and tagged . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s