4 ms·
Scalability is relative. To some people a 10TB database is large, but some teams I work with regularly see databases in the 10PB range. Going to that kind of sc
by klodolph 7y ago
Scalability is relative. To some people a 10TB database is large, but some teams I work with regularly see databases in the 10PB range. Going to that kind of scale you tend to see different design tradeoffs.
- pjmlp 7y agoThose kind of dataloads were already typical 15 years ago in the telecommunication domain, for example Nokia NetAct report processing for the whole operator network, based on HP-UX with Oracle OLAP support. Or the CERN infrastructure to deal with non-stop handling from accelerator data and storage for posterior data analysis. Sometimes this industry feels like a continuous reboot.
- ed_balls 7y agoI'm not sure how OLAP relates to CockroachDB. The use case for CockroachDB is quite simple. You have millions of applications that runs Postgres. A large chuck of them have to process more and more data which a single master cannot handle and there is no easy way of scaling. Any option you pick does require a lot of changes to the codebase. This is were CockroachDB should come in handy. The dream is to have a drop-in replacement for Postgres that would scale, so all SELECT .. JOIN still works.
- pjmlp 7y agoBecause that specific example made use of Oracle distributed cluster scalability, already 15 years ago. Postgres still has a couple of tricks to learn from big boys databases.
- klodolph 7y agoI was thinking of OLTP databases, with transactions, joins, random writes, and everything operating at scale. Running reports is much easier. I doubt CERN is dumping all of their experiment data into a single SQL database.
- pjmlp 7y agoBack in 2000 - 2005 it was in cluster of Oracle ones. Nowadays no idea.