Why Google’s Spanner Database Won’t Do as Well as Its Clone | Hacker News Reader

Why Google’s Spanner Database Won’t Do as Well as Its Clone

nextplatform.com

37 pointsKatydid23 comments

I'm used to reading inaccurate descriptions of Google and Googlers in the press, so I wasn't surprised by this from the article:

"... be usable as a single database for Google’s developers, who could care less about how a database or datastore is architected and implemented so long as it gets closer and closer to the SQL-based relational database that is the foundation of enterprise computing."

I don't think I've ever heard a less accurate description of Google engineers. The idea of implementing anything at Google scale on top of an enterprise-style SQL database would be vociferously shouted down by the vast majority of Googlers.

And the rest of the article is pretty much a mess without a discernible point.

> "... be usable as a single database for Google’s developers, who could care less about how a database or datastore is architected and implemented so long as it gets closer and closer to the SQL-based relational database that is the foundation of enterprise computing."

There's more than that worse with this quote. The whole thing is predicated on the concept that a "SQL-based relational database" is some sort of Platonic ideal of databases, and that all worthwhile databases must be an approximation of that ideal. This could hardly be farther from the truth -- there are plenty of successful NoSQL databases out there already, and one gets the sense that Spanner's SQL support is largely a compatibility layer on top of a very different underlying database.

Not even close to Google caliber engineer and I was also struck by that mis-characterization

This article doesn't seem to say why spanner won't do as well as it's clone.

"Its clone" -- what clone?! Spanner has only barely been announced. It's incredibly premature for an article to start making claims about a "clone" of Spanner, when we haven't even seen anything about Spanner yet beyond marketing information.

Spanner was announced in 2012; Cloud Spanner was announced this week. The clone is CockroachDB and has been in the works for a few years.

Wow, OK. The article didn't make that part very clear at all -- I assumed they were talking about a hypothetical future clone, not one that already existed. :/

CockroachDB is referred to as the clone here. They once gave a talk literally titled "Towards an Open-Source Spanner" (https://www.youtube.com/watch?v=ndKj77VW2eM), so I don't believe that's an unfair characterization.

It's CockroachDB. The guys who created Spanner just left (the clone) called Cockroach Labs.

The Spanner paper was published in 2012 on the Google Research site. It describes the entire system. We know a lot about the internals beyond marketing information.

https://research.google.com/archive/spanner.html

I think it's implied in the disappointment that Spanner isn't open source:

"To put it bluntly, it would have been more interesting to see Google endorse CockroachDB and support it on Cloud Platform, creating an open source community as well as a cloud service for its Cloud Platform customers."

CockroachDB will presumably do better in the author's eyes because it's open source and there can be a real community built around it.

Well that article was FAR too long then.

> But what would the world look like if the search engine giant had started selling capacity on its vast infrastructure back in 2005, before Amazon Web Services launched, and then shortly thereafter started selling capacity on its high level platform services?

What if Google did start selling capacity on its vast infrastructure in 2008, and we got Snapchat as a result?

> To put it bluntly, it would have been more interesting to see Google endorse CockroachDB and support it on Cloud Platform

To put it bluntly, no.

And he never even attempts to answer the question in his headline.

Honestly, I did not understand what this article talks about. I was expecting something along these lines (at least): https://quizlet.com/blog/quizlet-cloud-spanner

I can tell you why it will: stability. A post here a while back showed CockroachDB just recently started focusing on it where F1 RDBMS and Spanner have been running large workloads reliably at Google for some time now. I have high hopes for CockroachDB but Spanner has a lead here for now. Plus it's available so an ecosystem will start forming.

Keep in mind that once something is accomplished, it's not too long before other companies do the same as well. I see no reason why Azure/AWS can't do the same in next few years. If that gives Google an edge, great! But it won't be for too long.

I'm not sure, but it looks like maybe AWS already has... https://aws.amazon.com/rds/

You can't compare that to Spanner. RDS is a product that helps you manage/scale something like MySQL for you. It would compare to CloudSQL in the Google portfolio.

However, they have their own database engine called Aurora, which is what you actually get when you select MySQL or Postgres for RDS and provides much better throughput and replication behaviour. You can also only have up to 15 Aurora replicas whereas Spanner goes far far beyond that.

Aurora uses a MySQL-compliant API, but non-Aurora MySQL RDS instances are not using Aurora. Those instances are just managed MySQL. Ditto for Postgres (Aurora Postgres is in closed beta)

Generic RDS is just hosted MySQL. A better comparison might be Aurora:

https://aws.amazon.com/rds/aurora/

Still a very far-fetched comparison. Aurora is "just" a read replica system underneath a fairly vanilla MySQL. Better than MySQL, but miles away from the sophistication of Spanner.

It's true that copies or improvements can happen.

its more of a clickbait and doesn't talk anything about Google's clone. I wonder how it made to the front page of HN

Reply on news.ycombinator.com