CSE 141 Turn-in assignment number 1
Estimate the density of disks and tapes
Due Friday, Jan 25 in class
Important notice: This is an individual
homework. You should not help classmates, nor seek help from
anyone other than the instructor or TA's. (Exception - it's OK to ask
a classmate what the questions mean if it isn't clear. Still, it would
be good to check with me anyway.) In particular, do not even suggest
web sites or other sources for finding information to other students.
Each student should perform his or her own search for information on
the technologies, and figure out how to perform the estimates.
There is no single correct answer to the question. In fact, I'm
interested in seeing the range of answers that the class members come
up with.
Imagine that you want to set up a business that will provide
long-term storage of huge data archives, and need to decide whether to
use magnetic disk or tape technology. One issue -- the one we will
focus on here -- is how much space will the disks drives or tape reels
or cartridges take up, projecting forward to (possible) future
improvements.
Aside: Actually, this is one of the less important considerations,
compared to questions about cost, convenience, and especially
reliability. But let's pretend you plan to store the data in special
shielded bank vaults that cost thousands of dollars per cubic foot per
year, so space is an important issue.
We assume that the customers will probably never
need access their data, and you'll copy it to a newer disk or tape
every decade, so all we care about is the amount of space taken up by
the magnetic media. Thus, for tapes, you are concerned only with the
volume that a reel or cartridge of tape occupies, not the size of the
tape drive. For disks, I suspect the most compact technology is to use
ordinary hard drives (including the motor and electronics as well as
the platters) rather than removable disk packs.
Your assignment is:
- Find (either on the web or in published material) the information
you will need to estimate the three-dimensional volume required by
disks for three dates, one being within the last year or two,
another about 5 or 10 years ago, and the third 10 to 20 years ago.
It's important that your information corresponds to
technologies that are up-to-date but available commercially at the
respective times. For instance, don't use press releases about laboratory
prototypes - they are often years ahead of when the corresponding
commercial products are actually available.
For each "factoid" you use, you should indicate the source (web page,
book, educated guess, or whatever) of the information in enough detail
that I can check your figures. But I don't want to see reams of
information that you don't use in the calculations below.
- For each of the dates, calculate the "data density" (i.e. data
capacity to three-dimensional volume ratio) in megabytes per cubic
inch that can be stored on a hard drive. Be sure to always indicate
the units of your numbers (e.g. MB/drive, inches/foot, etc) in
your intermediate calculations as well as the final answer.
- Plot the points on a logarithmic scale to see if the
exponential rate of improvement has been roughly uniform.
- Making the totally bogus assumption that the exponential rate of
improvement given by your two most recent dates will continue exactly
the same for the next 50 years, estimate the data densities of disks
for the year 2050.
- Repeat the above four questions for tapes. NOTE: This will only
count one point (as opposed to the answers for disk, which count four
points), so it's OK to decide it's not worth your time. But you'll
probably learn something about tape technologies if you do it.
Please make your answers as clear as possible, but also keep them
short and to the point. I'll have a lot of papers to go through, and
don't want to spend time trying to figure out what various scribblings
on your paper mean. The example below (of something we did in class)
might serve as an example of brevity and clarity.
-
Factoids
- ScreenSize: 1,000,000 pixels/image (estimating a 1000x1000 screen)
- RefreshRate: 30 images/second (text, page 12 says "30 to 75
times per second")
- Data requirement: 3 Bytes/pixel (standard for "24 bit color")
-
Calculation
Bandwidth to screen: about 10^6 pixels/image x 30 images/sec x
3 Bytes/pixel
= 90x10^6 Bytes/sec
= 90 MByte/sec