dwolla-stress

Stress testing Dwolla's API. Let's see what happens when we try to make 1000 API calls (nearly) simultaneously.

This project inspired me to create arete, a load testing library, which dwolla-stress now depends on.

Get started

This is built with node.js. You'll need to have that installed!

Clone the project. Install depedencies:

npm install

Install mocha if you don't already have it:

npm install -g mocha

Put your API credentials into keys.js run all tests:

mocha

Or, run an individual test:

mocha test/accountInfo.js

Method of action

All the desired requests are made asynchronously and, in practice, are sent not quite simultaneously (seem to be within <=1 ms between each other), but pretty quick.

accountinfo-1000 - Begin profiling
accountinfo-1000 - Starting to make requests - 0ms (total: 0ms)
accountinfo-1000 - Sending request #0 - 0ms (total: 1ms)
accountinfo-1000 - Sending request #1 - 17ms (total: 18ms)
accountinfo-1000 - Sending request #2 - 2ms (total: 20ms)
accountinfo-1000 - Sending request #3 - 1ms (total: 21ms)
accountinfo-1000 - Sending request #4 - 1ms (total: 22ms)
accountinfo-1000 - Sending request #5 - 0ms (total: 22ms)
accountinfo-1000 - Sending request #6 - 1ms (total: 23ms)
accountinfo-1000 - Sending request #7 - 0ms (total: 23ms)
accountinfo-1000 - Sending request #8 - 0ms (total: 23ms)
accountinfo-1000 - Sending request #9 - 0ms (total: 23ms)
accountinfo-1000 - Sending request #10 - 0ms (total: 23ms)

As the responses come in, we note the time since the last response, the total elapsed time so far, and the response body:

accountinfo-1000 - Received response for request #0 - 493ms (total: 520ms)
{ Id: '812-742-3301',
  Name: 'Gordon Zheng',
  Latitude: 0,
  Longitude: 0 }
accountinfo-1000 - Received response for request #2 - 27ms (total: 547ms)
{ Id: '812-742-3301',
  Name: 'Gordon Zheng',
  Latitude: 0,
  Longitude: 0 }
accountinfo-1000 - Received response for request #1 - 9ms (total: 557ms)
{ Id: '812-742-3301',
  Name: 'Gordon Zheng',
  Latitude: 0,
  Longitude: 0 }

Notice that the responses won't necessarily come in the order which they were sent.

When all the responses have arrived, we spit out a report:

 === LOOK MA', STATS! ===
20 requests fired, of which we got back 20 successful responses (100.00% success rate)

Longest time between responses:  493 ms
Shortest time between responses:  2 ms
Average response time interval:  48.05 ms

Shortest response time:  520 ms 
Longest response time:  991 ms
Average response time:  750.8 ms

Writing tests

Writing tests is easy. We use the mocha framework for testing.

Create a new file in test/. Start it off by importing the keys file, arete anddwolla-node, in that order.

var keys = require('../keys.js');
var arete = require('arete');
var dwolla = require('dwolla-node')(keys.appKey, keys.appSecret);

// flag to false to test production API
dwolla.sandbox = keys.sandbox;
dwolla.setToken(keys.accessToken);

Then the test itself is simple to write. The arete module wraps the load testing functionality and overrides http and https so that it can force the max number of concurrent connections. Just call arete.loadTest, passing in a config object with the name for the test, number of times to make the API call, a bounded function containing the API call you want to make, and mocha's done() function as a callback.

describe('Account Info', function() {
    it('1000 Account Info requests in quick succession', function(done) {

        arete.loadTest({
            name: 'accountinfo-1000',
            requests: 1000,
            concurrentRequests: 100,
            targetFunction: function(callback) {
                dwolla.basicAccountInfo('gordon@dwolla.com', callback);
            },
            printResponses: false,
            callback: done
        });

    });
});