Partilhar via

Cloud Economics … You’re Fired!

  • Artigo

We just published a great paper on the Economics of the Cloud.  It does a very thorough analysis of current IT practices and gives some concrete ways that the Cloud (both Public and Private) can lead to IT cost savings.  Unfortunately, it also white washes one of the most important economic factors of the Cloud, namely:

Your company does not need you anymore!

Fortunately for you, dear reader, I am not saddled by the need to coddle you.  If you’re toiling away as some type of server administrator for some type of company right now, you need to wake up and wake up fast.  In a few years, your job will be gone!  It’s like Dean Pritchard when he tells the students they are expelled …

pritchard

Let it sink in.  Good.

So now that you’ve faced the inevitable, let’s dive in a little deeper.  According to the Economics of the Cloud Paper …

“Infrastructure labor costs. While cloud computing significantly lowers labor costs at any scale by automating many repetitive management tasks, larger facilities are able to lower them further than smaller ones. While a single system administrator can service approximately 140 servers in a traditional enterprise,8 in a cloud data center the same administrator can service thousands of servers. This allows IT employees to focus on higher value-add activities like building new capabilities and working through the long queue of user requests every IT department contends with.” The Economics of the Cloud, Microsoft, November 2010, page 4.

They are to have us believe that overnight, your email administrator is going to start “focusing on higher value-add activities”.  Like what?  You think you can just grab Mort the long time server administrator and send him out into the world to gathering requirements from users for additional IT projects?  Wake up and smell Mort’s cube – it aint gonna happen.

This is not easy for me either.  I have to sit in rooms full of Morts during the Cloud Briefings that I give to customers and basically give them this bad news, namely Microsoft/Amazon/Google do server administration better than they do.   To some it comes as a shock.  Microsoft??  Better server administration then me, Mort the long time blah blah blah administrator??  No way!  Hey, Mort: Yes way.  And, we don’t just do kinda better, we do way better.

And therein lies the  economic justification that I think that Economics of the Cloud soft pedaled. Take a look at StackOverflow for example. Here is how they are built.  Here’s their traffic.  And here’s how Spolsky sums it up.  So let’s start with StackOverflow as our baseline and  do what all good engineers do … double it!  5 servers becomes 10.  Let’s say each server has 8 cores.  We’re talking 80 cores to run the whole thing.  Let’s do what all good engineers do … double it again!  That’s 160 cores.  And, what does 160 cores per month cost on Azure you ask?

Let’s see … Extra Large Instance is 8 Cores and Guy Off the Street price for an Extra Large instance is $0.96/hour.  Not all months are the same length, so let’s say there are 750 hours per month.  That’s 20 extra large instances running for 750 hours … $14,400/month on compute to get 4x the current capability of StackOverflow.  And we can add in a couple of 50gb databases and 100’s of TB of storage and plenty of transactions, but you get the idea.

Wow!  That’s expensive!  (That’s what Mort says in the meeting when he’s sitting next to his boss and his boss’ boss).  Not so fast Mort!

What is that in Mort units? Let’s see.  Mort costs us ~$75,000 per year in salary, plus some amount in benefits, overhead, management, equipment …  according to some guy at MIT who definitely looks like he knows what he’s talking about, it’s like 2.7 time base salary for Mort per year in total costs.  Let’s say Mort costs us $200,000 per year or $16,000 per month.

Uh oh.  But, that means that 1 Mort costs us more than enough compute power to have 4x the capacity of a top 250 internet site?  That’s very bad news for Mort.

What should Mort do?  It’s easy.  Mort should start getting retrained now.  And, what should Mort get retrained as?  A developer?  Absolutely not.  Development is commoditized as a skill and is going no where but down …

code_4_food

The answer, dear reader, is Software Architect of course!  Take a look at what CNN Money thinks …

best_jobs

Software Architect is #1 in terms of growth potential over the next 10 years -- more than double the growth of guys who drill holes in rotten teeth and fill them up again for a living! Why?  I will have to go into the full “why” in another post, but let me ask you this … with India, Europe, Russia and even China coming online with development capability, who is going to design all those systems?  Who is going to design all these Cloud systems?   Software Architects that’s who!

So, Mort, buddy!  The die is cast.   Make it happen.  By my clock, you’ve got less than 5 years to complete obsolescence.

Comments

  • Anonymous
    December 08, 2010
    But someone has to run the Azure data centers.

  • Anonymous
    December 08, 2010
    @craig Very true.  But, I think the ratio cited in the Cloud Economics paper is correct, you go from about 1 person to 100 servers to 1 person to thousands of servers.  You're talking about a huge drop off in need over the next 10 years.

  • Anonymous
    December 08, 2010
    The comment has been removed

  • Anonymous
    December 08, 2010
    The comment has been removed

  • Anonymous
    December 09, 2010
    The comment has been removed

  • Anonymous
    December 09, 2010
    Geoff: Agreed.  I'm saying move the complexity from running and managing the servers to designing new an innovative applications to run on those servers.

  • Anonymous
    December 09, 2010
    And I'm saying you dont understand the complexity of the operations world, and this is not going to occur even in the slightest way you are predicting. Virtualizing the hardware is a trivial problem in ops, we have always virtualized the hardware by making junior people take care of it.  Now you want to take over the Junior jobs.  Have at it! All the rest of the work hasn't gone anywhere, and you won't be the ones to deal with it.  The same people who were dealing with it previously, will still be needed to deal with it, but it will be much harder. I've been writing operational automation systems about 12 years now, full time, and it's only getting harder.

  • Anonymous
    December 11, 2010
    US gvt. to cut 800 data centers ... http://bit.ly/fGu5MK  Many, many private companies to follow.  Make no mistake, it's already begun.