Google Spanner: First Globally Scalable Database With External Consistency

vu1986 writes with this bit from GigaOm: "Google has made public the details of its Spanner database technology, which allows a database to store data across multiple data centers, millions of machines and trillions of rows. But it's not just larger than the average database, Spanner also allows applications that use the database to dictate where specific data is stored so as to reduce latency when retrieving it. Making this whole concept work is what Google calls its True Time API, which combines an atomic clock and a GPS clock to timestamp data so it can then be synched across as many data centers and machines as needed." Original paper. The article focuses a lot of the Time API, but external consistency on a global scale seems to be the big deal here. From the paper: "Even though many projects happily use Bigtable, we have also consistently received complaints from users that Bigtable can be difficult to use for some kinds of applications: those that have complex, evolving schemas, or those that want strong consistency in the presence of wide-area replication. ... Many applications at Google have chosen to use Megastore (PDF) because of its semi-relational data model and support for synchronous replication, despite its relatively poor write throughput. As a consequence, Spanner has evolved from a Bigtable-like versioned key-value store into a temporal multi-version database. Data is stored in schematized semi-relational tables; data is versioned, and each version is automatically timestamped with its commit time; old versions of data are subject to configurable garbage-collection policies; and applications can read data at old timestamps. Spanner supports general-purpose transactions, and provides a SQL-based query language." Update: 09/20 17:57 GMT by T : Also in a story at Slash BI.

Comments

Sounds really great

[Cornwallis]

until Google decides to make it (and all your data) go away...

Re:Sounds really great

[Big Hairy Ian]

Sounds like they are trying to elbow into Oracle territory. As for Google ditching it I doubt they'd be so lackadaisical about peoples mission critical data as they are about a glorified rss aggregate (IGoogle).

Re:Sounds really great

[javilon]

They don't tell you where to get it. And no source available from the website. Before they go into oracle territory they need a product and a price tag.

Re:Sounds really great

[jellomizer]

I am sorry moderators if you find this fact unpopular but that is the point of Google development problems.
Google has a problem of pushing out new and innovative development stuff then a year or so, if it hasn't skyrocketed they will just kill the project.

As a developer you need to choose tools that you know will last, not something that will be here today and gone next week.

Re:Sounds really great

[Nerdfest]

They also give away the source on occasion ...

Re:Sounds really great

[Azure Flash]

I would mod parent up if I could because I would've said the same thing, in different words and with a capital letter at the beginning of my sentence. I am seriously sick of Google stabbing users in the back.

Re:Sounds really great

[Anonymous Coward]

This is a paper on a piece of Google infrastructure, not a product.

Take your ignorant trolling elsewhere please.

This is a great step for people who need it

[Chrisq]

Though not many people will need huge multi-centre databases it has cracked some of the big problems. Interestingly some of these don't appear to affect google's main business.

Spanner has two features that are difcult to implement in a distributed database: it provides externally consistent reads and writes, and globally-consistent reads across the database at a timestamp.

One of the issues with large distributed data systems was that reads at different nodes could retrieve data at a different (though consistent) state. I have seen this on google, a search shows a recent news item, then another doesn't show it again, before it finally covers all nodes and is generally available.

More to it than NTP...

[vlm]

Making this whole concept work is what Google calls its True Time API, which combines an atomic clock and a GPS clock to timestamp data so it can then be synched across as many data centers and machines as needed.

I'm guessing there's a little more to it than reinventing and installing ntp on your DBMS server. That little bit more is the actual interesting part.

Re:More to it than NTP...

[glop]

GPS/Atomic clock is better than NTP. It's a system to distribute time that will have a 400ns precision (probably a couple microseconds once you reach the actual servers in the data center).
If you use NTP or message passing you can't synchronize data centers more accurately than a couple milliseconds (assuming you have paths that are quite stable between them as transit time can be corrected).
So basically GPS/Atomic clock lets you synchronize 2 systems that are far apart more precisely and without having to make them communicate.
Note that Atomic clocks protect them from GPS outages, so they can really rely on the timestamps.

Protect data from governments?

[schwit1]

"Spanner also allows applications that use the database to dictate where specific data is stored ..."

Would this put data out of legal reach from the Patriot Act if the data were stored in Sveden or Grand Caymans?

Re:Protect data from governments?

[rhsanborn]

Sure, just ask Julian Assange, or Kim Dot Com.

And hopefully you have no assets or presence in the US. Because while they may not be able to easily get the data in the Grand Caymans, they'll be able to make the rest of your life hell.

Why semirelational?

[leandrod]

From the original paper linked at the summary post above:

Spanner’s data model is not purely relational, in that rows must have names. More precisely, every table is required to have
an ordered set of one or more primary-key columns.

OK, relational keys should not be ordered. But the fact that each table must have a key makes it a relation, at least in
principle, so Spanner at first looks like it is in fact more relational than SQL. Am I missing anything?

Re:Why semirelational?

[Anonymous Coward]

RDBMs don't require you to define primary keys, Spanner does, because it's evolved from key-value DB. If you'd read on to next sentence:

This requirement is where Spanner still looks like a key-value store: the primary keys form the name for a row, and each table denes a mapping from the
primary-key columns to the non-primary-key columns.

Re:Why semirelational?

[slapys]

That's why Google developed F1: The fault-tolerant relational database over Spanner. This database provides a traditional schema without named rows, and supports transaction-based relational SQL queries. Very interesting: http://research.google.com/pubs/pub38125.html

Re:Why semirelational?

[leandrod]

I do see how F1 is a fuller featured DBMS than Spanner, but it seems less relational, if it does not require that tables have
keys.

Tables without keys are not relations, so one could argue that Spanner is more, not less, relational than F1 and other SQL
systems.

Re:Why semirelational?

[leandrod]

Codd's original book

Not a Book, but an article. Granted, Codd published a book aftwards, but that was not original anymore.

required every relation to have a primary key, even if the key is composed of all the columns.

Precisely.

Not sure about what do you mean with relationships here. Relationships are a concept of ERDs, not of the relational model.

Re:Why semirelational?

[leandrod]

RDBMs don't require you to define primary keys

Yes, they do. The few of them, that is. SQL DBMSs do not, but they are not relational.

This requirement is where Spanner still looks like a key-value store: the primary keys form the name for a row, and each table denes a mapping from the
primary-key columns to the non-primary-key columns.

This makes little sense to me, because it describes not a key-value store — unless you consider the ‘value’to be all
non-primary key columns, which would stretch the definition of a key-value store —, but a relational database relation.

GPS tampering

[nstlgc]

What happens when governments decide it's time to tamper with or block GPS signals?

Re:GPS tampering

[jiriw]

They will use WiFi triangulation to be location aware instead? :P

It seems it's the GPS clock signals they want to use here. When those are dropped I guess they'll fall back on their own atomic clocks. It might be a little less accurate 'though.

From T*A:

Google’s cluster-management software provides an implementation of the TrueTime API. This implementation keeps uncertainty small (generally less than 10ms) by using multiple modern clock references (GPS and atomic clocks).

Nothing in there about GPS being essential. Just needs 'multiple modern clock references'.

Re:GPS tampering

[Anonymous Coward]

Then you'll be too worried about aeroplanes falling from the sky and ships running aground to be worried about database consistency, I guess.

Re:GPS tampering

[glop]

Atomic clocks would take a couple years to drift. So they have the time to hire some ninjas or lobby politicians to get this fixed before there is any impact.

Re:GPS tampering

[heypete]

What happens when governments decide it's time to tamper with or block GPS signals?

To what end? What possible purpose would that serve other than to interfere with critical systems like aircraft and marine navigation? (Yes, I know that aircraft and ships have backup means of navigation but it would still cause significant disruption.)

GPS certainly has other uses, precision timekeeping among them, and disruption of GPS would interfere with surveying, time transfer, and a variety of other functions. Pretty much any modern country uses GPS in some way and would suffer from disruption (it's for that very reason that the Russians have built GLONASS and the Europeans are building Galileo -- to provide some degree of independence and redundancy), so anyone looking to tamper with GPS signals would likely be shooting themselves in the foot.

Re:Huh

[LQ]

And in British English is means, roughly, a fool. http://www.urbandictionary.com/define.php?term=spanner

For a second there, I read that as Google Spammer.

[QilessQi]

...and I thought, these Android apps are really getting out of hand....

Possibly an unfortunate choice of naming

[blind biker]

I can imagine a lot of people misreading "Google Spanner" as "Google Spammer".

Re:Possibly an unfortunate choice of naming

[javilon]

Google spanner of code ?

Re:Huh

[someones]

but also spanning like spanning tree.

Re:Huh

[Anonymous Coward]

And in American English it means a wrench.