Moving From Oracle to CouchDB: Data Management at CERN

https://cloudant.com/blog/moving-from-oracle-to-couchdb-data-management-at-cern/

by mbroberg • 13 years ago

68 19 13 years ago

cloudant.com

benjarrell• 13 years ago

Surprised it is not everyone else's reason: Cost.

Regarding database dumps, I'm curious how/why data pump[1] is not what he needed.

1: Overview of Oracle Data Pump: http://docs.oracle.com/cd/E11882_01/server.112/e22490/dp_ove...

mattzito2• 13 years ago

Yeah, I'm not going to defend Oracle (or their pricing model), but in this case it sounds like the person who did the migration was just not an Oracle expert. Data pump is great for database dumps (and very fast), and SecureFiles is basically tailor-made for storing large amounts of binary data. There's a feature for schema versioning that would have allowed them to seamlessly run multiple versions of the same schema exposed to different sets of clients.

Oracle is vastly more powerful than pretty much any of the open source technologies, but it's eventually going to go away because of scenarios like the one described in the article. Tech folks are starting their careers on OSS technologies, learning on them, getting comfortable, and then when they stat working on something like Oracle, they aren't familiar with all of the capabilities under the covers, and migrate off of it to the more familiar platform.

And the cost, of course.

outworlder• 13 years ago

And Oracle Corporation's business practices, of course. They alienate their own partners.

rch• 13 years ago

As other comments have noted, direct costs were probably not the driving factor here. One place I worked benefited from great academic pricing on Oracle products, and I was happy to have Oracle as our go-to vendor. But people with the knowledge to leverage the capabilities you mentioned are expensive (which is fine), and don't always go beyond 'being a good DBA' to develop an intuitive feel for the domain (e.g. biology or physics)

rdtsc2• 13 years ago

Oracle is vastly more powerful than pretty much any of the open source technologies

Does it have multi-master replication like CouchDB? What about append-only storage of data (so can do live backup snapshots)? Or a REST-ful interface (so can directly use it from web clients, via a simple proxy)? A web based data browser and viewer like Futon?

Becuase those are very important features I like in Couch. Now I am not saying it knowing Oracle doesn't have those feature, as I don't know Oracle much (maye it does). But it seems to me that your claim how Oracle is a strict super-set of all the other database technologies out there is a bit hyperbolic.

mbroberg• 13 years ago

Assuming I understand this correctly -- and I could be wrong -- but I don't think that the pump thing would work for this use case. I think it's used for other applications. In this scenario, it needs to be to a DB next to these small agents. So spinning up an Oracle instance for each isn't really viable. Also, the agents aren't necessarily in labs that have the nice Oracle deal that CERN central has.

Create3• 13 years ago

Cost has nothing to do with it: they get it "free", as in US contribution in kind. And obviously, in the end of the day, Oracle profits much more.

http://openlab.web.cern.ch/about/partners/oracle

mje__• 13 years ago

"In the end, Zdenek had a single-server CouchDB installation on the order of hundreds of megabytes of metadata stored."

I think the title is a little inflammatory, no?

jarek• 13 years ago

"[I]t came down to storing lots of binary attachments that could be easily shared with other applications in the CMS. That was the main reason for the move."

Yeah this is a non-story. A better headline would be "Why a small application dropped Oracle for CouchDB". Compared with CERN experiments that generate petabytes per year, this is like publicizing Google's technology choice for its employee-list database. Not that ATLAS data goes into an Oracle database, but...

Where are the zealous HN title mods when you need them, eh?

fs111• 13 years ago

hundreds of MB? What? Use sqlite and be done with it.

pjmlp2• 13 years ago

Misleading title. It is only about a single person that used to work at CERN.

CERN has tons of developer groups, each with his own set of technologies.

shmageggy• 13 years ago

And furthermore, as some other commenters have noted, this is only for a tiny administrative database, not the massive data that's being generated by the projects. If they stored that in couchDB, then this would be news.

Create3• 13 years ago

http://home.web.cern.ch/about/updates/2013/02/cern-and-oracl...

http://ais.web.cern.ch/ais/presentations/eoug99/Implementing...

google:// Oracle CERN Technical training: available places in forthcoming courses

pjmlp2• 13 years ago

I used to work at CERN, what are you intending to prove with posting links about Oracle at CERN without explaining them?

Groxx• 13 years ago

Important note that I'm not seeing in other comments here: they started with both Oracle and CouchDB. This was basically a cleanup in favor of CouchDB, since running two different kinds of DBs is best avoided if possible for simplicity reasons.

knodi• 13 years ago

I wish they would have asked if there was anything about couchdb they don't like or have concern about.

Moving From Oracle to CouchDB: Data Management at CERN

Related Stories