Thanks:  0
Likes:  0
Dislikes:  0

1. ## interview question

Moor's law says that the power of PC’s is doubled every 18 months.
Each version of Microsoft OS works 1.5 times slower than the previous one.

Question: How often must Microsoft deliver a new version of OS so that users don’t see the effects of Moor's law?

------------------
It doesnÂ’t matter whether a cat is black or white so long as it catches mice

2. ## Re: interview question

...1962...?

------------------
-- Jean

You know your medical benefits are being reduced when the ER changes to "Drive-thru Only" service.

3. ## Re: interview question

Greycat: Extremely interesting problem. I haven't been able to solve it, but I'd like you to tell me if I'm following the right method.

This seemed like a variation of the differential equation problems we solved in the XI STD (eleventh grade). The only problem is the second piece of data about the MS OS being slower. I'm not quite sure how to factor that into the solution.

So what I assume is that the rate of change of versions of the MS OS are proportional to the differential between it's own speed and the speed of the processor at time t. Which makes business sense methinks!

Let Sp be the speed(proc) and So be the speed of the OS at time t months.

dSo & [So - Sp] ,[proportional to]
--
dt

Therefore,
dSo
-- = k(So-Sp) ; where k is a constant
dt

It follows that:
dSo
------- = k dt
(So-Sp)

On integration, we get:

log(So - Sp) = kt + c ; where c is the constant of integration

At time t = 0, So = S'o, Sp = S'p
c = log (S'o - S'p)

We assume that S'o = 2 * S'p ( Which means that at the start, the MS OS is capable of supporting a processor even twice as fast. Farfetched, but for the purposes of our mathematics, we must strike funny compromises!)

At time t = 18, Sp = 2*S'p --------(Moor's Law)

log(So - 2*S'p) = 18k + log S'p
18k = - [log (2*S'p) + log S'p]

log So - 2 log (2*S'p)
---------------------- = k
18
With this constant k and values for either So or Sp, we can determine the rate of change of the OS.

Like I said earlier, the problem remains: what to do with the fact that the MS OS is 1.5 times slower than its predecessor ?

Essentially that distribution will look like this:

Any hints ?

-Suresh

[This message has been edited by punekar (edited 02-03-2003).]

4. ## Re: interview question

<BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Originally posted by GreyCat:
Moor's law says that the power of PC’s is doubled every 18 months.
Each version of Microsoft OS works 1.5 times slower than the previous one.

Question: How often must Microsoft deliver a new version of OS so that users don’t see the effects of Moor's law?
<HR></BLOCKQUOTE>

This was an interview question?

Even accepting the two premises, did you assume that "doubled power" and "slower OS" could conteract each other?

In this scenario, are you assuming that Microsoft must deliver a new OS, in order to shield users from the increased power of the PCs? (Oh, the horrors!)

------------------
- Joe (strazzerjATaolDOTcom)

5. ## Re: interview question

To the question:

Well, a point has to be determined where computer speed and OS speed are on par. And from then on, wouldn't the OS have to be released twice every 18 months?

<BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Originally posted by jstrazzere:
This was an interview question?

Even accepting the two premises, did you assume that "doubled power" and "slower OS" could conteract each other?

In this scenario, are you assuming that Microsoft must deliver a new OS, in order to shield users from the increased power of the PCs? (Oh, the horrors!)
<HR></BLOCKQUOTE>

This reminds me of something one a Software Engineering professor once told us students, many years ago. He said something like this:

"A couple years ago, word users were annoyed that the cursor in Work (for DOS) disappeared if you hold a key down (Backspace for example), because the PCs were so slow. Then computers got faster and there was enough processing power for the application to display the curser. After this they got even faster, too fast for many of the users. So Microsoft added a GUI (Windows) and developed a new version of Word. And now (well, bakc then :-) ) the users are happy again, since the curser disappears in Winword when they hold a key down."

I thought that this was a pretty funny view of the things and is true to some degree. Of course, computers can never be fast enough for many different kinds of applications, but try to imagine how much idle time a 3GHz CPU would spend under DOS!

And from a cynic point of view... What else is the fancy eye candy in WinXP doing than slowing down the computer to user acceptable levels?

------------------
Convenience is the first step to enslave yourself. Freedom is inconvenient.

6. ## Re: interview question

If this were an interview question, it would tell me a lot about how the interviewer frames problems or thinks about the world.

Moore's Law, appropriately, is based on the number of transistors per square inch on integrated circuits. Moore's conceptual idea was based on the notion that this figure had doubled every year since the integrated circuit came on the scene. The prediction was that this was a trend that would continue and, as it turned out, this value of density, more than anything else, seemed to show a rough doubling around every eighteen months or so. So saying the "power doubled" is too simplistic. Saying that "each version of Microsoft's operating systems works "1.5 times slower" than the previous one is stated without contextual referrents because that does not just mean the chip capacity that the operating system is running on.

This also fails to take into account that, in fact, technologies using lower dielectric constant than the silicon dioxide insulating layer used most commonly up to now are actually starting to make Moore's Law invalid and this has been happening for awhile now, particularly as people look at variations on chemical vapor deposition and spin-on dielectrics (with carbon or silicon-based polymers). Thus if I was asked this as an "interview question" I would point out that the premises upon which it is based, based on how it was stated, are potentially misleading or, at best, highly approximate, and that degree of approximation would translate into any putative answer. Beyond that, I would say that the contextual referrents to determine Windows users "not seeing Moore's law" would require a further statement of delineation.

In other words, as the book "The Logic of Failure" adequately states, do not subsume multiple issues into one issue by asking about those multiple issues in the form of a single question. (I suppose one could make the argument that this is a good question if the interviewer was in fact trying to see how someone thinks about the issue rather than just what answer they come up with.)

------------------

7. ## Re: interview question

<BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Beyond that, I would say that the contextual referrents to determine Windows users "not seeing Moore's law" would require a further statement of delineation.<HR></BLOCKQUOTE>

Jeff, perhaps the qualification could be on the grounds of some kind of CPU benchmark. A rather crude one would be to run Ackermann's function on two processors with the same OS major/minor versions for a processing differential.

Then there are tests like the solution for hte linear system Ax=f, JPEG decoding, Zlib compression/decomp, text search, audio conversion, 3D vector calculation etc.

Ergo, the point of delineation would in effect be with "all things being equal" - with just a processor change, can the same processing time be achieved ?

The point you make hold valid to the extent that conventional CPU benchmarks - commercial or otherwise are not representative of typical business tasks/applications. Many of these have a high locality of reference and I can see a large L2/L1 cache making a substantial difference in observed performance. Modern processors are so tightly coupled with the platform architecture that it is extremely difficult to demarcate the limitations of tests.

-Suresh

[This message has been edited by punekar (edited 02-04-2003).]

8. ## Re: interview question

While I love this discussion, how much do you folks want to bet that if we got the person who asked the interview question to read this, that they wouldn't understand a word of it.

------------------
Scott Barber, Sr. Performance Engineer
sbarber@noblestar.com
http://www.noblestar.com
http://www.perftestplus.com

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts
•
Search Engine Optimisation provided by DragonByte SEO v2.0.36 (Pro) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.