October 2020

It’s not uncommon for different recommended practices to arise in technical sizing and optimization practices.  For many, it’s a compromise between most optimal data and ease of access vs. impact on production environments, which is no different from what we face when sizing Oracle on Azure.

overview

To continue on the UUID performance thread, I was recently asked by a customer how joins perform with various data types. I had not run a specific test, but suspected perform would be driven mostly by the size of the data types.

I wanted to verify my assumptions with real test data that shows the core performance of joins with CockroachDB.

the schema, data, and queries

For this test, two tables were created. The first table had one million rows and the second table had 200k matching primary keys for UUID, STRING, and INTEGER data types.

schema:

create table u1 (id uuid primary key);
create table u2 (id uuid primary key);

create table s1 (id string primary key);
create table s2 (id string primary key);

create table i1 (id integer primary key);
create table i2 (id integer primary key);

data load:

Publishing WordPress content on Twitter just got a whole lot easier. You can already bring single Tweets or threads, also known as Tweetstorms, over from Twitter to WordPress, and now we’re bringing the process full circle. With just two extra clicks, you can transform your entire blog post into a Twitter thread. By publishing your quality content on Twitter, you can open new lines of engagement and conversation.

A quick Google or Bing search and you’ll find no limit to the number of articles on databases that tell you that using “SELECT *” is a terrible terrible thing to do. Your code will be more fragile.  It is bad for your data dictionary.  You’ll end up with conflicts when you join, the list goes on. You can find a more reasoned argument from Markus Winand but ultimately for the majority of the time, the potential drawbacks outweigh any convenience benefits of using SELECT *.