Why Agent Swarms Will Break Traditional Databases | Jim Scharf

Jim Scharf (0:00)

Now, if you think about it, you basically either issue a request now, or more likely, you just have agents operating, looking for things and optimizing things over time, and they can spin up potentially tens, thousands of sub-agents, and they are in milliseconds, right? And so, things like scale and reliability of your database and your data platform and everything you built on top of it is gonna be even more demanding than today. So I think if you look back at today, you know, five years from now, you're gonna kind of laugh at the scale of which your business was operating from a transaction rate, particularly at the database level.

SPEAKER_2 (0:37)

Well, hello everyone, and welcome back to MTS. Today, I am joined by Jim Scharf, who is the CTO at MongoDB. Jim, super excited to talk to you today. I know you had a great keynote, and thank you so much for joining me.

Jim Scharf (0:49)

Yeah, thanks for having me. Yeah.

SPEAKER_2 (0:50)

What I think makes you particularly interesting, you've been working in the space for a really long time. You've been at AWS for 17 years, so over the last two decades, you've really seen the history of this large platform shift. Can you tell me a little bit about that?

Jim Scharf (1:06)

Yeah. Well, I mean, so I spent 17 years at AWS. I was there from before it was ever public, announced to the public, and then over the years, I found myself being pulled to implement key things that were unlocking growth for enterprises. Because when AWS first launched, it was kind of like these AI tools, it was cool, people just plunked down a credit card, and startup folks would start kind of get going, which was great, and it saw rapid early growth. But then at a certain point, AWS realized, hey, enterprise customers are like, hey, that's cute, but I can't also have someone launch an EC2 instance and also buy a flat screen TV with a corporate credit card. We need governance, we need control. So I got tapped to build what eventually became AWS Identity Access Management, which governed who could access what in the cloud, AWS Key Management Service, which controls the encryption keys, that protects customers' data in the cloud, and so forth. And so really, that's what I spent a lot of time at AWS. And then also really learned their large scale, mission critical services that can't go down. And some of our services that I operated at truly insane scales, and even just seconds of downtime could have blast radius, not just to other AWS services, but then to all the customers that depend on it.

SPEAKER_2 (2:24)

Can you tell me about that? I'm just particularly curious. So what are some of those services? I know if AWS goes down even for a half a second, it just completely messes everything up.

Jim Scharf (2:33)

Yeah. So one of the services that no one much thinks about is you make an API call to any AWS service, S3, EC2 or what have you, there's a service that authenticates and authorizes that API request. And I think since I've left, they've shared that service processes over a billion requests per second. Oh my gosh.

SPEAKER_2 (2:54)

So it goes down for a second, you miss a billion requests. Yeah.

Jim Scharf (2:57)

So it is really high scale distributed system operating, and it operates in all AWS regions worldwide. So that team there is great. I had the privilege to lead and grow that org from literally zero to what it is today. I left in 2022

SPEAKER_2 (3:12)

Yeah. Well, something that's interesting as well, I mean, 17 years ago, I'm assuming when you're building out this infrastructure, you didn't really realize how it would be used so much by agents. And often the times you're thinking about humans, you're building for humans, you're building infrastructure for humans. Very clear when it comes to identity, when it comes to permissions, when it comes to certain security principles. But that changes a lot in the agentic era. What do you think is one of the biggest areas where you see that changing?

Jim Scharf (3:38)

Well, I think one is scale for sure. I don't think we're come close to feeling even the earliest onset of the scale. Yeah, so when cloud first launched, everything was available by APIs. And so instead of humans clicking buttons, you could have a shell script doing a loop and doing things. So even from that human to scripted, automated, we saw over the years in cloud that scale went up rapidly.

Now, if you think about it, you basically either issue a request now or more likely, you just have agents operating, looking for things and optimizing things over time, and they can spin up potentially tens, thousands of sub-agents, and they are in milliseconds, right? And then if that starts cascading and they start making calls, Coinbase is saying, hey, what about your agents making, purchasing things and now Coinbase is getting calls? Yeah, so I think we're really in the earliest days. And so things like scale and reliability of your database and your data platform and everything you built on top of it is gonna be even more demanding than today. So I think if you look back at today, five years from now, you're gonna kind of laugh at the scale of which your business was operating from a transaction rate, particularly at the database level.