Hacker News new | past | comments | ask | show | jobs | submit login

It’s pretty clear that it’s a new frontend talking to a very old (probably green-screen style mainframe application) backend through some kind of gateway, though. Maybe with some modern DBMSes serving as ancillary data storage, but with the legacy system definitely being the system-of-record that everything has to interface with in the end.

If a new-ish web app for a bank or government service has a “maintenance period” where it becomes inaccessible between 5PM and 9AM each day—without the actual web layer going out, just seemingly changing its access ACLs on the data layer—that’s because they’ve got the web app fronting a mainframe that was designed for explicit OLTP ingest / OLAP batch ETL phase separation. The new record event stream gets collated at night to build queriable/indexed tables. Often on hardware so slow that it needs the entire maintenance period to finish building those tables+indices. (This is why better performance in these systems is hard: if you want higher OLTP ingest throughput on the same hardware, you can have it, but then they’ll need longer maintenance windows, so long that they’ll now break their SLAs by being unavailable during business hours!)




Thanks, I found this post fascinating and informative.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: