14 Comments

Summary:

Michael Stonebraker has helped create a lot of database systems and has lots of thoughts about where the industry is headed. Oracle, SAP, NoSQL, NewSQL — they all have a place if they play their cards right, but some might have a harder time than others.

Michael Stonebraker is a lightning rod in the database world. He knows his stuff, having helped create some of the most-popular database systems around — Postgres, Ingres and Vertica. One of his most-recent endeavors (among many), VoltDB, is an in-memory OLTP (online transaction processing) system he claims is two orders of magnitude faster than legacy options.

Yet, Stonebraker’s opinions throughout the years have been fairly controversial, garnering support on one hand and vehement opposition on the other. In 2011, for example, he told me that Facebook was trapped in a MySQL “fate worse than death” — and, boy, the backlash came fast and furious.

Stonebraker came on our Structure Show podcast this week to talk about his assessment of the database market as it stands today, including the fate of NoSQL, Oracle and, yes, Facebook’s MySQL installation. Here are some of the highlights, but if anyone into building, using or investing in database technologies will probably want to hear it all. (P.S. It’s a longer episode than usual, with Barb Darrow and I talking about the news for the first 18 or so minutes.)

Download this Episode

Subscribe in iTunes

The Structure Show RSS Feed

One size will fit none

“In every vertical market I can think of, there’s some better way to solve your problem than using a legacy relational database system,” Stonebraker said.

He has actually been preaching this idea for a while, but now it seems more legitimate than ever. There are columnar architectures for analytics, in-memory architectures for transactions, and, of course, NoSQL architectures for simple key-value operations and new data types. Even graph databases are catching on commercially.

Want proof of how mainstream these new types of databases are becoming? “Obamacare, for better or for worse, is being launched on a NoSQL database system,” Stonebraker noted.

There’s room for a lot of winners

“There will be three or four or five, or maybe six, categories of database systems architected very differently, in each one of which there will be two to three successful vendors,” Stonebraker predicted. “And I think the core, meaning the legacy relational database system, is going to slowly shrink. This all will play out over perhaps a decade.”

Source: Wikipedia Commons / dcoetzee

Source: Wikipedia Commons / dcoetzee

NoSQL will come back down to earth

“My prediction is that NoSQL will come to mean not yet SQL,” he said.”… “Cassandra and Mongo have both announced what looks like, unless you squint, a high-level language that’s basically SQL.”

The perceived value of a purely low-level language all but gone, Stonebraker thinks NoSQL systems will also come to embrace ACID capabilities. It might already be happening.

“I think the biggest NoSQL proponent of non-ACID has been historically a guy named Jeff Dean at Google, who’s responsible for, essentially, most to all of their database offerings. And he recently … wrote a system called Spanner,” Stonebraker explained. “Spanner is a pure ACID system. So Google is moving to ACID and I think the NoSQL market will move away from eventual consistency and toward ACID.”

Oracle will feel the squeeze from SAP

“The other thing I think that’s just real fascinating that hasn’t really gotten much press yet is that SAP is in the database business and that … SAP customers are Oracle’s biggest customer right now,” Stonebraker said. “Among the elephants, there’s going to be a duke-it-out between Oracle and SAP.”

It’s a little early for this to really happen yet, and we don’t yet know how SAP’s customers will respond to any attempted persuasion to switch databases, but, he added, “My expectation is that SAP will make a compelling case for their … customers switching off of Oracle and onto HANA.”

Facebook will keep searching, possibly fruitlessly, for a MySQL replacement

“Facebook has one of the hardest data management problems on the planet,” Stonebraker said. “They have been trying for several years to move off of MySQL and onto something else, and so far they haven’t found anything else that can do the problem at the scale they need done.”

It’s a step back from his stance a couple years ago, perhaps based on some of the information Facebook has shared on its MySQL efforts since then and how well it has continued to hold up. However, it’s still not an endorsement of MySQL as much as it is a recognition of Facebook’s database chops.

Generally speaking, Stonebraker said (falling back on a familiar, but funny, turn of phrase), “The code bases the legacy vendors are selling right now are 25 years old, and it’s time for them to be retired and sent to the home for obsolete software.”

Feature image courtesy of Shutterstock user semisatch.

  1. Interesting read about the future of DB’s… if anything this shows that there’s a whole set of new DB technologies out there trying to solve the next big problem and THE solution is sort of in a state of limbo right now.

    Share
  2. NoSQL databases have certainly enjoyed the ‘cool new kid’ hype for a while, and they will inevitably come down to be evaluated realistically – but the truth is, SQL is old. Graph databases will generalize well into quantum hardware and beyond; long term, my bet is on those.

    Share
    1. You’re already betting on quantum hardware? That’s the most intriguing thing I’ve heard in a while :-)

      Do you have a more elaborate blog post about this opinion, that you could link to?

      Share
  3. Really, the best argument seems to be “SQL is old” or “RDBMSes are old”. So is C and Unix. Sometimes the reason old stuff sticks around is because it was actually a pretty good idea.

    Share
    1. Well said.

      Share
    2. Agreed!

      Can’t stand this kind of argument.

      Share
    3. No, the best argument is that standard RDBMS’s get shoehorned into tons and tons of roles that they aren’t the best option for. I’m extremely proficient with SQL and I’ve been a DBA, but the fact of the matter is that relational databases aren’t that great for a lot of the tasks for which they are routinely used for today. (Data Warehousing is an obvious example, and don’t even get me started on how awful they are for analytics) Yes, they are fantastic for anything requiring ACID, but guess what? There’s a lot of stuff out there that doesn’t need it.

      If you know your data upfront, and are pretty sure there won’t be any major changes down the road, then go relational. But if you need to do exploratory analysis on large datasets, and need to structure your data as you learn it, you are wasting your time with relational. There are much better options out there.

      Share
  4. Frankly the assessment above is very generic and being a db user myself not something new.

    Share
  5. michaelhausenblas Tuesday, December 17, 2013

    I have to agree with kk above, not too much surprising here (heck, the one-size-fits-it-all meme/paper is from 2005). But I do admire Mike and his achievements and if you really wanna see him in action and learn something from him I suggest to check out a recording called ‘Stonebraker Live! 2013 – Cambridge’ which is available via YouTube: http://youtu.be/j2q927rUXPM

    Cheers,
    Michael

    Share
  6. Michael Stonebraker’s opinions and claims are always refreshing to read. He’s done a lot for our industry and for how we do data processing. Some of his claims are certainly right as well.

    However, as any software vendor, the other half of his claims should be read with caution. Today, the most popular DBMS (relational or not) are still Oracle, MySQL, and SQL Server. Even his “popular” PostgreSQL is still a niche player, let alone the almost forgotten Ingres and the never really popular Vertica. Obviously, I’m not saying they’re *bad* databases, but they’re certainly not very *popular*. The same goes with SAP. Their Sybase databases have been surpassed by SQL Server both in quality and in popularity 10 years ago. I hardly believe that Oracle and Sybase will have the “final fight” for RDBMS supremacy.

    But again. That’s Mike Stonebraker, the salesman as in “The Traditional RDBMS Wisdom is All Wrong”:
    http://blog.jooq.org/2013/08/24/mit-prof-michael-stonebraker-the-traditional-rdbms-wisdom-is-all-wrong/

    ;-)

    Share
  7. “The other thing I think that’s just real fascinating that hasn’t really gotten much press yet is that SAP is in the database business”

    This is news to me. What database is he referring to?

    Share
    1. SAP HANA,, plus there’s Sybase IQ which now comes under their wing

      Share
  8. Ddear masters of the future of databases: please make it look like an open source Datomic.

    Share
  9. Conclusion: Facebook needs to hire Jeff Dean, at any cost.

    Share

Comments have been disabled for this post