Performance Testing

Aspects Tested

• User querying stocks (due Nov. 6th)
• User favoriting stocks (due Nov. 6th)
• Critical path (due Nov. 13th)
  • See how vertical scaling helps
  • Test on medium-large DB (~10,000 records)

Goal

• Test performance of website by having an "average" user do something normal

Prep work

• call scripts/warm_cache.py to warm up the cache
• Generate 2,500 user accounts
• Each account should have ~4 favorites
• There should be 1 (or a few) stocks that aren't favorited, but cached

Instance Configurations

• Follow these instructions to keep a consistent testing environment
• Use this chart to see how many CPUs are available
• Set ProcessParallelism to be the number of available CPUs
• Leave each process single threaded
• Make servers for M3Medium, M3Large, M3XLarge and M32XLarge using our custom SinglePassengerENV.json or MultiPassengerENV.json files
• If making a load balanced server, have 2 instances running and use the same instance type for the database (db.<instance type chosen for server>).

Script

• User logs in
• User visits dashboard
• User visits dashboard for a stock not favorited
• User favorites stock
• User visits dashboard for favorites
  • randomize the order?
• User unfavorites the stock it favorited during test
• User logs out

Clean up

• None

Aspects to Test

User querying stocks

• Should hit current price, historical prices, and sentiment data (graph)
• Need to test with cache hits and cache misses

User favoriting stocks

• Favorite stocks
• Unfavorite stocks

Critical path

• User logs in
• User queries a stock
• User favorites that stock
• User queries a favorited stock
• User unfavorites that stock
• User logs out