2 ms·

Getting Rid of RDBMs

5 points by dedalus 19y ago

corentin 19y ago
I don't get it. I don't design systems including databases, so I may not be qualified to speak (I've studied information system design a bit, though). What's wrong with relational theory and Codd normalization rules? It looks like a good way to organize data and keep it coherent. Just because it's 30+ years old isn't a reason to reject it!
neilc 19y ago
What a terrible article. Leaving aside the grammar and spelling mistakes, the quote from Stonebraker is taken completely out of context (in the interview, he is talking about the architecture of systems for doing algorithmic trading, not relational databases in general: latency is indeed critical for algorithmic trading, but far less so for typical database applications and web apps). Then there are content-free statements like: "This means you should be paying attention to your code quality, optimization ... There are built in limitations as dictated by clear and proven underlying mechanisms that prohibit current modern database and application technology from scaling much further." More broadly, I don't agree with the idea that if something makes sense for the Amazons and Googles of the world, it should be adopted for run-of-the-mill web applications. Amazon and Google can afford to write tons of infrastructure from scratch and hand-tune it to match their performance and scalability requirements; the typical web app developer cannot, but then again, the typical web app has totally different requirements. Simply generalizing from the top end of the cost vs. scalability tradeoff downward doesn't make a lot of sense. Doing joins and referential integrity in the application layer has real costs -- discarding 30 years of technology overnight has real tradeoffs that the article doesn't even bother considering. The article does have one redeeming feature: the linked-to interview with Stonebraker and Margo Seltzer is worth reading. http://delivery.acm.org/10.1145/1260000/1255430/p16-stanik.htm?key1=1255430&key2=3880943811&coll=&dl=ACM&CFID=15151515&CFTOKEN=6184618 http://delivery.acm.org/10.1145/1260000/1255430/p16-stanik.h...
tx 19y ago
The question remains in the air: how do I perform more complex queries without SQL? The proposed distributed hash table won't do it. The article mentioned eBay's scaling PDF offering a link at the bottom. I followed the link read every page of that PDF and have not found any evidence of them abandoning (or minimizing use of) RDBMS. eBay still runs Oracle, but uses in-house built index for listings search, but that's hardly a solution for general problem of running queries more complex than SELECT * FROM Foo WHERE ID=n. Besides NOBODY uses SQL to do complex full-text searches, even small companies I worked for. We used Lucene. I would love to have a "distributed SQL": scaled-out SQL server in proper sense of the word, capable of running complex SQL queries on distributed sets of data. Hopefully by the time we grow to eBay size someone will innovate in that regard. Or, perhaps, it exists already?

5 points by dedalus 19y ago

I don't get it. I don't design systems including databases, so I may not be qualified to speak (I've studied information system design a bit, though). What's wrong with relational theory and Codd normalization rules? It looks like a good way to organize data and keep it coherent. Just because it's 30+ years old isn't a reason to reject it!

neilc 19y ago

What a terrible article. Leaving aside the grammar and spelling mistakes, the quote from Stonebraker is taken completely out of context (in the interview, he is talking about the architecture of systems for doing algorithmic trading, not relational databases in general: latency is indeed critical for algorithmic trading, but far less so for typical database applications and web apps). Then there are content-free statements like: "This means you should be paying attention to your code quality, optimization ... There are built in limitations as dictated by clear and proven underlying mechanisms that prohibit current modern database and application technology from scaling much further." More broadly, I don't agree with the idea that if something makes sense for the Amazons and Googles of the world, it should be adopted for run-of-the-mill web applications. Amazon and Google can afford to write tons of infrastructure from scratch and hand-tune it to match their performance and scalability requirements; the typical web app developer cannot, but then again, the typical web app has totally different requirements. Simply generalizing from the top end of the cost vs. scalability tradeoff downward doesn't make a lot of sense. Doing joins and referential integrity in the application layer has real costs -- discarding 30 years of technology overnight has real tradeoffs that the article doesn't even bother considering. The article does have one redeeming feature: the linked-to interview with Stonebraker and Margo Seltzer is worth reading. http://delivery.acm.org/10.1145/1260000/1255430/p16-stanik.htm?key1=1255430&key2=3880943811&coll=&dl=ACM&CFID=15151515&CFTOKEN=6184618 http://delivery.acm.org/10.1145/1260000/1255430/p16-stanik.h...

tx 19y ago

The question remains in the air: how do I perform more complex queries without SQL? The proposed distributed hash table won't do it. The article mentioned eBay's scaling PDF offering a link at the bottom. I followed the link read every page of that PDF and have not found any evidence of them abandoning (or minimizing use of) RDBMS. eBay still runs Oracle, but uses in-house built index for listings search, but that's hardly a solution for general problem of running queries more complex than SELECT * FROM Foo WHERE ID=n. Besides NOBODY uses SQL to do complex full-text searches, even small companies I worked for. We used Lucene. I would love to have a "distributed SQL": scaled-out SQL server in proper sense of the word, capable of running complex SQL queries on distributed sets of data. Hopefully by the time we grow to eBay size someone will innovate in that regard. Or, perhaps, it exists already?