NoSQL database

[nomadd]

Some of us were quite good at that stuff back in the day.

Several TB of data, all memory resident, spread across a 20 node, 80 CPU, 640 core cluster - with guaranteed sub millisecond access to any piece of data, including complex aggregations?

That really was some impressive "stuff" you were doing "back in the day."

[OwlHoot]

Sum bozo was on 'bout this today. What's the crack with NoSQL databases; google and others use 'em? Do I need to worry?

(if there is a solution to the problem don't worry - if there's no solution to the problem don't worry)

Yours in peace.

They're potty to use[*] - I got the hang of MongoDB in half an hour and was merrily coding away using it in perl & python scripts

[*] But for flip's sake don't tell the clients that!

[eek]

They're potty to use[*] - I got the hang of MongoDB in half an hour and was merrily coding away using it in perl & python scripts

[*] But for flip's sake don't tell the clients that!

+1. Mongodb is easypeasy.

[swamp]

NoSQL:

1. Little or no normalisation or type safety.
2. Data structures have to be designed in terms of how you retrieve them.

Now point 1. may be a bit scary for old-school developers/DBAs, but with vast numbers of automated tests you can argue you don't really need guaranteed DB integrity.

Point 2. sounds awkward, but on the flip side you can change your schemas whenever you want.

Mix it all together and you get something that is massively scalable.

[Sysman]

Several TB of data, all memory resident, spread across a 20 node, 80 CPU, 640 core cluster - with guaranteed sub millisecond access to any piece of data, including complex aggregations?

That really was some impressive "stuff" you were doing "back in the day."

Ahem.

For the hardware available 30 years ago we got some impressive results.

Says Sysman who managed to cut something that took all weekend to run down to 20 minutes...

By redesigning keys.

My statement still stands.

[oscarose]

NoSQL:

1. Little or no normalisation or type safety.
2. Data structures have to be designed in terms of how you retrieve them.

Now point 1. may be a bit scary for old-school developers/DBAs, but with vast numbers of automated tests you can argue you don't really need guaranteed DB integrity.

Point 2. sounds awkward, but on the flip side you can change your schemas whenever you want.

Mix it all together and you get something that is massively scalable.

Tis interesting. Apparently, traditional RDBMS couldn’t handle the vast amount of data the likes of Google processes. I’ll be learning these new techniques over the season of good will and applying to the portfolio – a new box is on order to conduct these tests.

[swamp]

Tis interesting. Apparently, traditional RDBMS couldn’t handle the vast amount of data the likes of Google processes.

Facebook used to run on MySQL. It didn't scale so they wrote Cassandra instead.