Online transaction processing

It's only transaction processing.

It's only OLTP.

And then it's any kind of transaction processing, whether it's financial or non-financial.

It could be counting at scale.

Yes, I can agree.

Thanks.

I guess I should also maybe backtrack and clarify, you know, Tiger Beetle is only OLTP, so it doesn't replace OLAP or OLGP.

You still have Postgres for all your strings, you know, your customer names.

And this is also reflective of the growth that PRDB and ClickPipe saw, right? Like from five customers to a thousand plus customers moving terabytes from Postgres to ClickHouse.

So what we thought was like, this is becoming the de facto like default data stack for the world, right? A lot of companies, right? Why not we now manage both sides of the stack, which is Postgres for OLTP and ClickHouse for analytics and bring them closer to to make developers' life super easy, right? So what we are doing, I'll talk about the product specifically.

At a high level, what we're doing is we are integrating these two technologies, right? Like we are making all the workflows to integrate these two technologies very easy for developers.

Let it be like moving data, querying data, right? Like all of that will be now magical.

That is number one.

Second is what we did, Michael, was like the approach that we took for Postgres is offer Postgres backed by local NBME storage, right? So we are going back to basics.

we're going back to basics, right? Where we believe that for OLTP, tail latencies matter a lot, right? Like every like millisecond or like even half a millisecond matters, right? Because you look at the concurrency and you look at the number of requests, say a Salesforce, like you open the dashboard, like it's millions of requests, like that single dashboard can make, right? Like not across like hundreds and thousands of users, right? So every millisecond matters, which is why we took the approach where for Postgres, we have a architecture where storage is co-located with compute, which is NVMe back storage, right? And obviously it has all the platform features like availability, backups, reliability, security, which I'll touch upon.

But it really gives best possible performance for customers.

To do a simple DAX query, it has to do, you know, 50 joins in the back between all these tables and try and understand how they all relate, and your DAX becomes stupidly complex because it has to then pick out all these tables in different places.

So you need to denormalize and actually turn it from an OLTP, or online transactional processor model, data model, into an OLAP, tabular model, which is just dimensions and facts.

And there is rep- repetition and redundancy, and that's okay because it's just a structure that Power BI could work with and it can aggregate over, analyze it.

So that's, that's data model.

And that translation from a business use case to an engineer was all about contextualization of the problem and having the engineer understand to some extent be responsible for like, okay, I understand software and data best practices.

I understand OLAP versus OLTP databases.

I understand like when it should be relational versus JSON structured.

I know how to write Python code and how to use pandas.

experience opinion, a strong opinion, I would say, is that you cannot have read replicas with off.

For general OLTP case with many users, and I say this after a lot of experience, I built three social networks with each one of them achieved many million users and one plus DAO daily active users.

And we needed to scale reads, of course, right? So we needed read replicas.

First one I built in 2006.

But you need to do it on the primary as well, right? Because who knows what happens there? It's no difference

here.

I think the big difference is people not offloading just like OLTP traffic to a replica.

I think it's people think that I've got a replica, I can send my reporting or analytics queries there, or I can give a data team access to that, and it doesn't matter.

First on the technical level.

So he actually goes and he provides a diagram of the architecture of how you would use the different data stacks from the pre AI era when you were just using an OLTP database and an app to Then you adopt a cloud data warehouse in 2016.

You start building a cloud-first data function.

Around 2020, you bring in DBT, you bring in Fivetran, and all of a sudden, all of your data is in the same place and you're able to actually build a reporting system where you're able to answer most of the questions kind of horizontally, vertically across your business.

Absolutely.

For the audiences not familiar, OLTP is essentially transactional.

OLIP is essentially analytical.

And there is no certain benchmark I would put, but there are certain criteria I would look for.

So one of the examples I would look for is when you're looking for latency budget of milliseconds, you want to stay in OLTP

world.

But when your latency budget is in about seconds and minutes and you're audience is mostly analytical, training the ML workloads on data, and it's not dependent on transactional data, I would move towards the OLAP.

would make more sense.

And listeners should weigh whether this reflects genuine enterprise demand or aggressive partner incentives from AWS.

Sub-200 millisecond latency uses AWS EC2 placement groups as a notable technical claim for OLTP and ERP workloads.

The host may want to scrutinize what conditions and configurations are required to actually hit the number in production.

There's a zero ETL integration with Amazon Redshift and direct access to Amazon Bedrock SageMaker and Quick Service lets customers apply AWS AI tools to Oracle data without moving it, which is a pretty good selling point.