[FoRK] Multicore, async segmented sequential models

Stephen Williams sdw at lig.net
Thu May 9 11:15:07 PDT 2013


On 5/9/13 9:32 AM, Aaron Burt wrote:
> On Thu, May 09, 2013 at 08:39:54AM -0700, J. Andrew Rogers wrote:
>> MongoDB has been invaluable when teaching database engine design. It
>> allows me to tie specific naive design elements you would never expect to
>> see in real systems to real-world database meltdowns and failures. It
>> plays Goofus to the rest of the industry's Gallant.
> But a very charming Goofus.  Who seduces naive coders with sweet promises
> of data persistence that's even simpler than ActiveRecord, and like,
> totally scalable and stuff.  The Ops guys keep saying it's "web scale" and
> giggling, so that must be a good thing, right?
>
> Performs well on Powerpoint slides,

Is there a Powerpoint slide powered cloud somewhere?  Or just Excel?

At least it is written in C++ rather than Haskell (STIG, acid-state, and I think one more).

This is an apparently complete list of NoSQL databases, although they don't cover content stores like S3 and similar:
http://nosql-database.org/

PostgreSQL seems to have been making a name for itself in the geosearching area.  Useful even if all you use it for is an index.

On the MySQL side, MariaDB has been getting press.  And I've been spammed about NuoDB, not to mention various fully commercial 
MySQL scalers.

Cassandra is well supported in some cloud packages, as is S3 compatible content stores.

For fine-grained work, I'd probably favor something like Neo4J.

Savvy real-world comments on the best of these for different kinds of tasks would be very helpful to avoid recapitulating the 
pain of wasted exploration.  Most things work well enough in a prototype, and most fail in annoying ways at scale.  These days, 
we ought to have a rich development and data model, very simple install and management from embedded, single instance, to 
scaled, redundant, replicating, etc.

>    Aaron

sdw



More information about the FoRK mailing list