The options aren't just give $10B to AWS, Oracle, IBM or Microsoft. The JEDI should have been a multi-award IDIQ (indefinite delivery indefinite quantity) contract. With a multi-award IDIQ the each task order issued under the IDIQ is competed for among the multiple vendors the IDIQ has been awarded to. It is actually fairly unusual that this type of IT services contract was not done as a multi-award IDIQ.
Having to compete for each task order is thought to reduce costs to the government. Of course having to compete each task brings its own additional costs which don't necessarily make the cost saving proposition true. The big benefit I see to this contract is that it helps prevent vendor lock-in. I don't worry about the next 10 years and $10B, but what happens after 10 years when the DoD is now locked in to proprietary AWS cloud features which make migrating to a different vendor technically and fiscally near impossible.
You should not do infrastructure as IDIQ, this is an infrastructure contract. The idea that all contracts issued by IT are for "IT Services" is part of the reason digital systems at the DOD and other government agencies are such a nightmare. HAving a single system to build off of is the best way to do it, and not having that system be a customer solution for just the government is way better then not.
Lock in is the only concern, but from my experience doing an IDIQ or multi-vendor won't solve the lock in problem. The best thing to solve that is to make sure your using bland normal things not customer government thigs.
You know, I've been in some data centers and top 50 HPC clusters ... I don't understand why the US government doesn't run it's own cloud systems, except for the appropriations. There are enough computer scientists in the larger org to field a major distributed systems engineering group.
Now, think of the political climate: there’ve been decades trying to “trim waste” by steering money to the private sector, where efficiency is taken as a matter of faith. First you’d need to hire the technical leadership to properly plan something like that and then hire all of the people who’d build it, not to mention procuring all of the hardware & data centers, etc. You’re looking at years before you have a positive return on any of that – and all the while, every big contractor is going to have their lobbyists talking about how it’s a waste of money compared to going with them.
NASA tried this a decade ago with Nebula, which was part of the early OpenStack history. I don’t know exactly what happened with that but the people I know working on NASA projects are all using AWS.
I'm glad you mentioned the caveats with IDIQ contracts. I would say that the IDIQ process rarely, if ever, results in savings. (In fact, if you have some examples where it did I know some people who would be interested in studying those cases.)
The history of IDIQ processes is likely a large part of the reason people were anxious to do something different in this instance. Now that it's all politics, it'll probably cost more and function less when all is said and done.
Some aspects of the military, I don't really miss.
I’ve done IDIQ contracts in state government. You can absolutely save money, but you need to force a limited list of SKUs and renegotiate pricing at different volume-based high water marks.
Winner take all is also important, because it eliminates the room for sales bullshit. The magic is figuring out how to break up the lots of services so that it makes sense. You don’t want to give any one of AWS, Google and Microsoft everything, so you want to design lots that keep it competitive. This also ensures that any one vendor doesn’t subsidize strategic SKUs by discounting something that isn’t meaningful.
Everything fed is more complex, but it’s all doable.
What’s your theory on the justification DoD is giving to go single-award IDIQ?
I doubt their actual justification will ever be public, but I suspect their motivation is that they require a common IaaS platform such that they can have a unified toolset and a single, comprehensive approach to infrastructure security.
Keep in mind also that fewer proprietary features at AWS are available when considering standardizing across all classification types. “East/west” and GovCloud don’t have feature parity.
Especially since that 10B becomes 20B when in 5 years, Oracle tells you that you that your initial implementations are being EOLed and you need to start over.
I've been torn on this contract from a professional experience POV. In the IC we got AWS after they beat out IBM and the IBM protest. I'm an AWS fan so I was excited and it worked well for me when I was a contractor (I'm no longer in the space).
But having the DOD on AWS along with the IC concerns me from a lack of diversity standpoint. I've always said that Microsoft and Google should have teamed up a bit to make a competitive run. But I don't think anyone was ever interested in sharing the contract to begin with.
Azure has a large presence in the government and they have a fairly robust offering that multiple agencies use. So its not as big a mono culture as you might think. I know this one contract makes lots of noise because its the DoD and people think the military is this huge organism, but the Department of Veterans Affairs was on tract to spend 1B on cloud per year and they use Azure ( as well as AWS )
It blows my mind a department dedicated to Veterans Affairs can spend 1 billion on cloud computing in a year, do you have a source or any more information about what they’re spending that on?
They're probably the largest hospital provider in the country. Looking at their wiki it says there's 152 hospitals and 1400 outpatient facilities. Plus all the other VA responsibilities they fulfill.
It's a big organization and they have a lot of 'customers'
Because you’d rather have wasteful spending than moderately productive spending...? Seems like you want the government to waste money just so you can say “I told you so”.
Like it or not, defense spending is a huge boost to the middle class stem job market, and taking it out would have consequences wider ranging than you think. With “white collar welfare” I’d argue the gov gets a better return on its investment than most other options, while arguably making it harder for developing nations to become a threat via brain drain because the stem market is strong here.
Yes, precisely—wasteful defense spending accomplishes the goal of supporting the middle class with jobs while failing at the goal of killing people. It's not an "I told you so" thing. I'm genuinely fine with massive taxpayer-supported boondoggles that create jobs in the long term. My objection to defense spending is that it kills people, not that it takes money from my paycheck to create more paychecks.
As you may remember from the Project Maven controversy, there are a lot of people on HN and in the tech community who are fundamentally opposed to the US military being more effective at carrying out its missions.
