My Retail

Installation instructions

Notes: Tested on OS X with Ruby, managed by rbenv. Homebrew installed native dependencies, e.g. readline, postgres (tested with 9.5.4)

brew install readline

Install ruby. Tested with 2.4.1. rbenv used to manage Ruby versions. Install bundler gem, which will manage all other gems installed.

bundle

Once the application gems are installed, running rake will run the build, and should be an additional verification point that everything is set up.

Create databases

There are a number of ways to do this, one way is from a psql prompt, run:

create database my_retail_dev;
create database my_retail_test;

To create the database schema:

rake db:migrate

To load data into the database using the Rails seeds functionality:

rake db:seed

Confirm data is loaded (example, may change):

# \c my_retail_dev
You are now connected to database "my_retail_dev" as user "andy".
andy@[local]:5432 my_retail_dev# select * from products;
 id | external_id | price_details |         created_at         |         updated_at
----+-------------+---------------+----------------------------+----------------------------
  1 |    15117729 |               | 2018-02-15 18:19:19.761739 | 2018-02-15 18:19:19.761739
  2 |    16483589 |               | 2018-02-15 18:19:19.765197 | 2018-02-15 18:19:19.765197
  3 |    16696652 |               | 2018-02-15 18:19:19.767898 | 2018-02-15 18:19:19.767898
  4 |    16752456 |               | 2018-02-15 18:19:19.770103 | 2018-02-15 18:19:19.770103
  5 |    15643793 |               | 2018-02-15 18:19:19.772462 | 2018-02-15 18:19:19.772462
(5 rows)

e.g.

 select price_details->'current_price'->'value' as value, price_details->'current_price'->'currency_code' as currency from products where external_id = 13860428;
 value | currency
-------+----------
 13.49 | "USD"

Overview

The Product resource has an ID and price details, stored as JSON data, using the jsonb column type.

API Testing

Fetch product details

Using Postman or curl, make the following HTTP requests.

GET http://localhost:5000/products/{id}

e.g. 15117729

GET http://localhost:5000/products/13860428

Response (e.g.):

{
    "id": 13860428,
    "name": "The Big Lebowski (Blu-ray)",
    "current_price": {
        "value": 13.49,
        "currency_code": "USD"
    }
}

Curl request

curl -X GET http://localhost:5000/products/13860428

Updating product price

Make a put request to the following URL with the following JSON body as the payload.

Set the body to application/json.

PUT http://localhost:5000/products/13860428

{
    "product": {
      "id": 13860428,
      value": 17.50
    }
}

Curl request

curl -X PUT \
  http://localhost:5000/products/13860428 \
  -H 'content-type: application/json' \
  -d '{
    "product": {
    	"id": 13860428,
    	"value": 17.50
    }
}
'

Production Curl Request examples

curl -X PUT \
  https://limitless-oasis-33699.herokuapp.com/products/13860428 \
  -H 'content-type: application/json' \
  -d '{
    "product": {
    	"id": 13860428,
    	"value": 17.29
    }
}
'

curl -X GET https://limitless-oasis-33699.herokuapp.com/products/13860428

422 Response example

curl -X GET https://limitless-oasis-33699.herokuapp.com/products/123

Running the Application

To boot up the application on port 5000:

rails s -p 5000

Or:

foreman start

Benchmarking

Benchmarking with wrk tool https://github.com/wg/wrk

$ wrk -t1 -c10 -d10s http://127.0.0.1:5000/products/13860428
Running 10s test @ http://127.0.0.1:5000/products/13860428
  1 threads and 10 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency   135.66ms   85.83ms 753.22ms   95.61%
    Req/Sec    40.25     11.22    50.00     78.26%
  382 requests in 10.08s, 176.82KB read
Requests/sec:     37.89
Transfer/sec:     17.54KB

External Integrations

HTTP interaction with Redsky service

Additional Performance Improvements (future)

Cache the Redsky service calls periodically, TTL of 1 minute

As requests for product IDs come in, we could cache the response with a key of the product ID, and the value as the JSON string, e.g. in memcached or another key-value store, and then for subsequent requests for the same ID, server the JSON string from the cache store and not make a live HTTP call.

We could specify a period of time that the cached response is fresh, e.g. 1 minute (the TTL).

Optimize JSON serialization

Benchmark various JSON serialization options.

Consider a slim API response variant

Using a parameter, a client could request a reduced set of attributes that represent the object. For example, in a mobile app on a summary screen, a trade-off for a faster summary screen could be fetching fewer attributes for each row, which would take less time to serialize, and produce a smaller JSON payload.