Often you think of a solution to a simple problem and once you come up with that solution you realise you need to apply this to a large dataset.   In this post, I will explain how I deployed a simple solution to a larger dataset while preparing the system for future growth.  Here’s the state of play before changes in my client’s eCommerce system:

Existing System:

  1. Login to aggregator’s portal to retrieve datafeed URI
  2. Login to customer admin interface to create or update merchant details
  3. Create cron job to pull data from partner URI after inital setup
  4. Cron job dumps data in MySQL
  5. Client shopping UI presents search field and filters to customers to search and use
  6. Search result is extremely slow (homepage: 2.03 s, search results page:35.73 s, product page:28.68 s ).  Notice the search results and product pages are completely unacceptable

Proposed System:

Phase 1:

  1. Follow steps 1-4 of existing system
  2. Export MySQL data as csv
  3. Create an instance of Elasticsearch with an index to store product data
  4.   Export MySQL data as CSV
  5. Create script that bulk insert the exported data into Elasticsearch
  6.   On command line, search Elasticsearch instance using various product attributes (product name, type, category, size etc.).  Check the time speed of search results.

Phase 2:

  1. Build a search interface that uses Elasticsearch
  2. Display search results with pagination
  3. Add filters to search results
  4. AB Test existing search interface and Elasticsearch based and compare conversion (actual sales)
  5. Switch on the best solution – Elasticsearch

A few libraries already exists that can solve some of these challenges e.g

  • Elasticsearch-CSV – https://www.npmjs.com/package/elasticsearch-csv
  • SearchKit – https://github.com/searchkit/searchkit
  • PingDom – https://tools.pingdom.com

In the next post, I will dive deep into how I used Elasticsearch-CSV to quickly ingest merchant data and the response I got