1 |
On Tuesday 02 March 2010 19:07:21 walt wrote: |
2 |
> On 03/02/2010 04:23 AM, Arttu V. wrote: |
3 |
> > On 3/2/10, walt<w41ter@×××××.com> wrote: |
4 |
> >> This article was a big surprise to me. Am I the last one to hear about |
5 |
> >> this stuff? |
6 |
> >> |
7 |
> >> http://news.cnet.com/8301-13505_3-10461670-16.html?part=rss&subj=new |
8 |
> >> s&tag=2547-1_3-0-20 |
9 |
> > |
10 |
> > If you're expecting a discussion then perhaps you'd care to narrow it |
11 |
> > down a bit: which part of the article are we expected to feel |
12 |
> > surprised about? |
13 |
> |
14 |
> I was surprised that three major social networking sites have dumped |
15 |
> MySQL (but now the article says only two sites). I've also not heard |
16 |
> of the "NoSQL" movement before, and I'm curious to know what's motivating |
17 |
> it. Maybe nobody trusts Oracle? |
18 |
|
19 |
Because Codd's relational database model, as implemented by Oracle, Sybase, |
20 |
PostgreSQL, MSSQl and a slew of others, is not the only way to model a data |
21 |
storage system (aka database). In much the same way that a bakkie with a |
22 |
canopy is not the only way to transport workers, as buses do exist. |
23 |
|
24 |
Relational databases are demonstrably mathematically correct, but like all |
25 |
things they have their limits to how far they can scale. More often than not, |
26 |
this limit is imposed by how fast the db engine can access and identify data |
27 |
using the hardware upon which it is built. Traditional RDBMSes don't even |
28 |
vaguely scale to the levels Facebook runs at. |
29 |
|
30 |
The NoSQL movement is nothing more than an effort to find other ways of |
31 |
extracting data having consciously ditched SQL for the job. By way of example |
32 |
(this is not NoSQL per se, it illustrates the point), Google's data extraction |
33 |
methods are not even remotely SQL. Heck, they aren't even completely correct, |
34 |
they are merely "good enough". See what happens when you dump the old mind-set |
35 |
and look at fresh new ideas? Oftentimes you get something that works better |
36 |
than the old way. Google does not care that their search results are not 100% |
37 |
spot on, they are good enough for your query. If other stuff that they missed |
38 |
deserves to be higher in the ratings, it will climb higher over time till it |
39 |
does show. Considering the size of Google, this is a very workable approach. |
40 |
|
41 |
|
42 |
|
43 |
-- |
44 |
alan dot mckinnon at gmail dot com |