 I'm a little bit passionate about, I hate databases. My team knows this about me. I hate them from a certain company in particular, and I will not name that name, because it could be a sponsor. Corona and I don't want to get you in trouble. I really want to say the name, and I'm not going to. So they are, to me, slow, monolithic, expensive. They run 24-7. They require people to tune them. They require all of us to work in this horrible language called SQL. We're dealing with Edwin Codd's normalization from the 1970s, and it was all great when it started, but time has just passed it by. And so I focused a little bit on the cost, and I produced, actually, we didn't produce the next graphic, but I threatened to so many times that my team will recognize this immediately. This is Patrick. He would work here with us, but you chose a database in designing your application. Patrick hopes you like your database. For us, this is a reality that some of the databases that we've chosen in our solutions are 50, 75,000 euro a year when you consider the server, the license, and all of the overhead that goes along with it. Which would you rather do, add another person to your organization, or keep something like this running 24-7? That, for us, has been an important thing to work through. And I'll give you an example of where a database and KVS solution came together. So we have a whole bunch of clients, I've vastly simplified this, who have us map where their business happened. So one of the things you see is some different fields here across the top. And they're basically saying, well, last year, Atlanta was part of the South region. Now it's part of the Southeast. Can you remap that and give us some insight into our sales trends? Sure, we can do that. But we have these maps changing constantly. Now the kind of thing is you're taking in 100 million records. You say, OK, I'll create a lookup table like this by clients, and I'll just hit it. It's doable in SQL. It makes all the sense in the world, except that you're running 100 million records past it every day. You're going to have huge latency issues that will kill you. So for us, it was important that we put these into a schema. But we faced this and said, SQL's just not going to cut it. Converting this to key value stores allowed us to scale this out and use Hadoop and do this very, very quickly without any of the bother. And then frankly, going back to the ephemeral, the servers go away. There is no database that's standing up and holding this forever. You shut it down. Cost is zero. So we'll move on.