I work for a gambling company and the bet-pool/wager-resolution process happens with another company category called a 'tote'. For 49 of the US states we work with one particular company because we've identified them as being overall more competent and accurate. For the remaining state, regulations require that we work with a company with a 'local presence' in the state. There's only one company that fits that boundary, and they are somewhat awful to deal with. Not only do they constantly have problems, but they're not very responsive with business interactions. A good portion of our application infrastructure is dedicated to just handling the special case for this one state.

The state thinks it's doing good by forcing a monopoly for this one company with 'local presence', but the hilarious thing is that the company is HQ in Europe somewhere, and our preferred partner is actually US based.

What state if you dont mind me asking?

New Jersey. There's a long running joke that it's an interference point for the Mob.

Is it possible for you to share the state that does this?

I agree that a single supplier is always bad. It is basically walking into a customer lock-in situation.

However, among the major cloud computing providers Amazon is definitely one of the best in terms of offerings, maturity, and stability. Microsoft is not far behind Amazon, but I do not have any experience using Azure to know how it compares.

For "military-grade" cloud computing, I would choose Amazon.

I agree, but under the condition that controls and governance are put into place that ensure the DoD builds their infrastructure and applications in a way that allows them to move to another provider with little friction (this is less about being multi cloud and more about ensuring that the technology choices you make don't marry you to the provider).

We should not be headed down a path where AWS becomes "too big to fail" or considered a systemically important institution (for obvious reasons). Today's underdog is tomorrow's incumbent.

Pretty sure that is not possible because if this contract is anything like the previous 2 that AWS has won, they will literally be building entire airgapped datacenters for the government full of their custom server designs running all of their custom cloud software. You're not going to be able to just uninstall dynamodb, EBS, or S3 from the datacenter servers and expect to be able to spin up one of Azures or GCPs matching service using the same hardware.

Why would you not expect to be able to port your data and applications to another cloud provider? They all can run x86 virtual machines, they all can store data on iSCSI equivalents (EBS) or in object stores. No one is writing custom code to run on custom hardware at cloud providers (except perhaps to optimize for specific hardware extensions, GPUs, etc).

Sure you can probably port your code over just fine, but it's going to take 3 years for another cloud provider to build an entire airgapped datacenter and launch each internal service in it. Considering the CIAs 600 million dollar contract got an entire one dedicated for them, I wouldn't be surprised if this contract which goes up to 10 billion dollars ended up involving multiple new regions. As I said, another cloud provider won't be able to simply just take over those datacenters and install their software, they'll probably have to build their own airgapped datacenters which will take years.

So no, there is no way it's going to be easy for them to port over to another cloud provider.

If the DoD makes it a requirement that a new service provider can take ownership of the physical facility and it’s contents, a new provider could most certainly assume responsibility for the cloud system. It’s a contractual issue, not a technical issue, no different than China requiring AWS’ China data center region be operated by a Chinese company versus AWS proper.

I've worked for multiple major cloud providers and I assure you that launching the dependency chain to get most cloud services in production for 1 region takes 6 months minimum (they have entire teams dedicated to launching regions). That is after building the datacenter (years) and launching the main database services (which require custom server hardware as I previously mentioned) that power most of the main cloud services.

Google uses spanner, AWS uses dynamodb, and Azure uses Cosmodb which are all different database systems and thus they are all completely incompatible.

I hope that Amazon forces the US Government to renew the contract at a 1000% markup. Don't like it? Too bad, single source, good luck with your migration, we've already amortized the equipment. Oh, you only have X months to migrate before we stop patching security vulnerabilities? Well, you better get to migrating quickly!

Whoever thought this was a good idea, shame on them.

That's not how these contracts actually work. Certainly Amazon COULD act against their customer's interests but the customer will be around for a long time, so that would be really short-sighted.

The process for re-bidding them starts a long time before the contract ends, and often the incumbent will offer a temporary contract to cover the time between the end of the contract and the beginning of the next contract. The process is unpredictably long because one of the losers -- especially if it is the incumbent -- may protest the award.

A similar thing happened with a contract I worked a lot with and the new contract (much smaller than JEDI, but still ~$2BN). The contract was to replace legacy datacenters with cloud services and service management layered on top, and was an IDIQ (Indefinite Delivery, Indefinite Quantity) contract -- it was held up in protest so long after award that the government just ordered nothing, and came up with alternate strategies. The protesting entity was one of the winners of the bid (unlike JEDI, it was awarded to 4 parties, though they were all reselling AWS, Azure, Google, etc) and they ended up with nothing from this very expensive and time-consuming bidding process.

Why would they do that? Amazon is no match for the US government and military complex and pissing them off is the worst business move they could make.

10B over 10 years is really not that massive a contract. It would end up as something like 1/8th of AWS revenue, though more at Microsoft’s cloud division.

It’s also less than 0.2% of all federal government contracts, and assuming it’s the only contract they had it would not even place them in the top 50 government contractors. http://www.fi-aeroweb.com/Top-100-US-Government-Contractors....

I understand your point, but there is an entire makret for government contracting which is basically a monopsony. Difference complience, etc. But how much money can we really justify to deal with the hassle of a cross-cloud expense? Also, more providers = woder attack surface. The likelihood that at least one of two or three companies will screw up their security is greater than the likelihood that one of one will.