Google and Facebook Team Up to Modernize Old-School Databases

http://www.wired.com/wiredenterprise/2014/03/webscalesql/

by salemh • 12 years ago

48 25 12 years ago

wired.com

mikkelewis• 12 years ago

Currently under what circumstances will we see performance increases, and by how much?

jpalomaki• 12 years ago

Previous discussion in HN at least here: https://news.ycombinator.com/item?id=7480843

Project page: http://webscalesql.org/

rakoo• 12 years ago

I don't think "Webscale" is the right word; they should have used "Worldscale". The web part is irrelevant here.

peferron• 12 years ago

It's a joke referring to the "MongoDB is Web Scale" meme. The FAQ is quite suggestive:

Q: Why is it called WebScaleSQL? A: While there are a variety of origin stories for the name depending on who you ask, ...

They just don't want to mention it directly. :)

RyanMcGreal• 12 years ago

Every time I see the word "Webscale" I think of this animation:

http://www.youtube.com/watch?v=b2F-DItXtZs

coldtea2• 12 years ago

The name was SPECIFICALLY picked to refer to that animation, it's an inside joke kind of thing.

mixmastamyk2• 12 years ago

What is the rationale of putting this work into Mysql? With its sloppy execution and the dark clouds of Oracle hanging over it? Why not Postgres, or even MariaDB instead?

davidw2• 12 years ago

Because they are locked in to Mysql and can't easily switch to something better.

MrBuddyCasino2• 12 years ago

I'd really like to see the evidence for that. Postres > MySQL, and MySQL doesn't have that many lock-in features last time I looked.

Any Google/FB/Twitter employees here than can elaborate?

nl• 12 years ago

Sounds like a fork of MySQL, which makes it similar to MariaDB (Note that Google's performance patches for MySQL are similar already).

morgo• 12 years ago

Facebook has previously answered this question on Quora: http://www.quora.com/MySQL/Is-Facebook-considering-ditching-...

Also from the WebScaleSQL FAQ:

Q: Why didn't you base this on MariaDB, Percona Server, Drizzle, etc.... A: We reached a consensus that MySQL-5.6 was the right choice for this, as it has the production-ready features we need to operate at scale, and the features planned for MySQL-5.7 seem like a fitting path forward for us. We will continue to revisit this decision as the ecosystem evolves.

collyw2• 12 years ago

So relational databases are old school, and column stores are all the hip trendy thing, despite throwing away a ton of features?

jlouis• 12 years ago

Well, there are three types of databases: Relational, Columnar, and finally massively distributed. They throw away different feature sets in order to handle a certain subset better.

The primary reason Relational databases are so friggin' strong is that they have a solid foundation and maturity. Most new "NoSQL" stuff is immature crap which should never be used in production for any kind of persistent data. But they are being used as such, and I do note that matures the products over time.

collyw2• 12 years ago

Tell me about it. I have been hauled into a project, where we are reimplementing a relational database using NoSQL. No idea why, other than the fact that they want to use NoSQL.

ams6110• 12 years ago

We are now clearly in a world where many technology journalists are entirely unaware of the history of computing before the year 2,000 or so.

coldtea2• 12 years ago

Worse, we are now clearly in a world where many developers are entirely unaware of the history of computing before the year 2,000 or so.

I mean, we had "NoSQL" before we had SQL. Tons of schema-less, hierarchical etc DBs, talking directly to your program and such. Anybody that uses them without massive scaling needs doesn't know his history.

avz• 12 years ago

Relational databases are not the right choice for large-scale data stores serving large number of users because:

- they provide unnecessary features at high cost (e.g. database-wide transactions),

- they don't provide features that are essential (e.g. scalability, distribution, graceful degradation).

By contrast NoSQL data stores like Bigtable are highly scalable and can easily be stacked over to provide more features if needed (see e.g. Megastore, http://pdos.csail.mit.edu/6.824-2011/papers/jbaker-megastore...)

arrc• 12 years ago

Well not just Google / Facebook but Linkedin and Twitter are also in the team.

Google and Facebook Team Up to Modernize Old-School Databases

Related Stories