CRM Perf

A demo CRM for my talk on "The Thick Front-End".

There are 5 applications in this series:

CRM Rails. A standard Rails app with no ajax and an API. Shows the baseline of how most apps are developed today.
CRM Rails-jQuery. A fork of CRM Rails that uses basecamp-style ajax.
CRM Angular. An Angular.js app that interfaces to either CRM Rails or CRM node.js.
CRM node.js. A drop-in replacement for the CRM Rails API written in node.js.
CRM Scala. A drop-in replacement for the CRM Rails API written in Scala with Play.

Back-End Performance Results

DO NOT READ TOO MUCH INTO THESE. This was just me playing around writing super contrived applications. You should try these experiments (and more!) yourself to determine what the right stack for your application is. I'm not trying to prove a point about anything here, just having fun.

Results

Framework	Database	Notes	Hits	Response Time
Rails (Webrick)	PostgreSQL	HTML	5814	199ms
Rails (Thin)	PostgreSQL	HTML	8256	177ms
Rails (Webrick)	PostgreSQL		5722	278ms
Rails (Thin)	PostgreSQL		6525	431ms
Rails (unicorn 3x)	PostgreSQL		16060	252ms
Rails (unicorn 4x)	PostgreSQL		13439	315ms
Rails (puma/jruby)	PostgreSQL		3437	512ms
Rails (Thin)	MongoDB	no orm	4875	560ms
Sinatra (Thin)	MongoDB	no orm	12455	173ms
Sinatra (puma)	MongoDB	no orm	8585	234ms
Sinatra (unicorn 10x, 1000 users)	MongoDB	no orm	28215	42ms
Sinatra (unicorn 10x, 2000 users)	MongoDB	no orm	46286	205ms
Sinatra (unicorn 10x, 4000 users)	MongoDB	no orm	65494	152ms
Sinatra (unicorn 20x, 5000 users)	MongoDB	no orm	65907	291ms
node.js	PostgreSQL		26080	100ms
node.js (1000 users)	MongoDB		27304	71ms
node.js (2000 users)	MongoDB		46064	198ms
node.js (4000 users)	MongoDB		52390	246ms
Scala/Play	MongoDB		25811	68ms
Scala/Play (4000 users)	MongoDB		84012	60ms

Notes

anything not marked as 'HTML' was hitting api/v1/customers.json.
node.js tests never used an orm
It's very interesting that rails + mongo was slower than rails + pg since you'd think mongo would be faster, and that there was no ORM with mongo. This was repeatable too.