Hacker News Reader

Comment by iamEAP | Hacker News Reader

As an end-user of databases, I hope in 50 years time, there will be no meaningful distinction between OLAP and OLTP DBs.

We spend so much time and energy copying data around into specific data stores to solve specific problems / answer specific questions / enable specific features... It's messy, complex, and adds a ton of overhead.

It could simplify a lot of technology if an operational database could also handle non-trivial analytical workloads.

View on HN

This would be amazing, but it's an open question in my mind if it's possible to have a system that works sufficiently well for both OLTP and OLAP use cases at large scale. HTAP systems today are still very raw, and it intuitively feels like we'll be bumping into the Pareto frontier, where optimizing for analytics use cases can only come at the expense of transactional performance.

You’re not going to run your analytical queries off your production DB once you have more than a couple users writing ad-hoc queries. And you probably don’t won’t to replicate your salesforce/ads/clicks into your prod db. So you’re going to create a replica. So why does it matter if the replica is a different type of database than prod?

Do any current databases support replicas with heterogeneous indexes? It seems like that would greatly reduce this problem.

Google Cloud Spanner kind of does what you want.

Isn't this the promise made by SAP HANA?

Oracle Active Data Guard also (queriable replica), with column store.

Reply on news.ycombinator.com