Unless things have changed, DoD is Microsoft's biggest customer. It's not that they're stuck. Between the combination of sheer scale (not thousands, millions) security requirements, and desire for a stable ecosystem, there's few, if any, that can match what they provide. This isn't a dig at others products, let alone "open" solutions. They've been serving DoD for decades, and you can count on a Microsoft product to come with solid security documentation (often with NIST 800-53 reference) , FIPS 140-2 compliance, PKI support (absolute must in the DoD Common Access Card environment), and(relatively) solid vulnerability management. DoD requires ALL of those things.
I wouldn't argue that they don't have room to improve. But that doesn't discount what Microsoft has accomplished.
So, this is a $927,000,000 five year contract. Right?
So let's tell the government not to pay talented companies to provide services and instead hire federal employees to build out infrastructure to do the same thing.
Assuming we have kind of a $200,000/year base salary for the types of people we need (so like a GS-15 or so), how much does that cost the government over the same time span?
$1,000,000/employee.
So you would prefer the government hire 927 people full time for five years to completely replace MS? Do you think they'll do a good job? Is $200,000 salary a good deal, or will you end up hiring bad software developers at that rate? Can you hire better developers for $400,000/year? Can 464 people build a MS replacement in 5 years?
Also, who will support it? Do we need to factor that in as well? Will we have to replace this new OS/tech stack in 5 years? Is that another billion dollars?
I suspect that you're looking at the price and being blown away by the number. The DoD spends $580 billion each year (roughly). Are there any comparable companies with a $580 billion/year run rate? Expand your mind, friend. This is a bargain price for the government.
You should do some digging and try to find out how much DISA is paying for their own version of AWS. I suspect that number will blow your mind as well.
I work on a small team of about 12 within 18F (part of the General Services Administration). We are creating cloud.gov, a Platform as a Service (think Heroku, Google App Engine, IBM Bluemix) for the Federal Government. The key is to remove the government compliance burden from federal government development teams while also making modern technology accessible and understandable.
We come from private sector for a two year civil service. We are super lean and all of our work is open source (github.com/18F).
Note that cloud.gov is only seeking certification for FISMA Moderate impact level, so a lot of DoD systems (such as anything classified) cannot be hosted on our PaaS.
This is a REALLY hard problem to solve and the government does not make this process easy. If we miss milestones, we don't get paid, so that $1,000,000 in salary that your team was paid for the last 90 days of work suddenly doesn't get paid. Your $100,000 in profit you forecasted suddenly turns into a $1,100,000 loss for that quarter.
When I started working in government way back in the early 00s, I too thought "I can just automate all these worthless idiots out of a job." If it was that easy, someone would have figured it out by now.
Most companies have to spend $30k - $50k just writing a proposal in the hopes they'll win some work. The problem at this size and scale is astronomical.
If you can think of a better way, I really would love to talk to you. Email is in my profile. :)
Two questions (on a tangent):
1. Define "qualified"
My experience, before I gave up on the DOD sector, is that companies that post Landry list reqs actually mean it. They are even worse than the larger industry at only interviewing candidates who look stellar on paper.
2. How much are you paying?
The few times I did get past the resume filter the salary was slightly below market, and none of these companies have anything resembling non-salary compensation.
120WPM * 60minutes/hour * 16hours/day * 5workdays/week * 52weeks * 5years * 50people / 10words/line
= 748.8M lines of code
Good luck!
:)