API V2.0 - Performance testing

[srtalbot]

Link to previous testing

Link to rough estimates of current submission numbers from yearly to minute values.

User patterns

There are two primary factors that impact system throughput:

1. How frequently does the client need/ want to pull the data? (Frequency)
2. How much data do they need to pull? (Volume)

Based on these factors, here are three user patterns to consider:

1. (high or low frequency, low volume): Clients pull data as it comes in, in low volumes
2. (low frequency, high volume): Clients that have significantly higher volumes that want to bulk download during one or two sessions
3. (high frequency, high volume): Clients that need to pull large volumes regularly throughout the day for a longer period of time.

Acceptance criteria

We can discuss threshold, but using these as a starting point

• [ ] Dev and Product to review load testing limits
• [ ] Identify how many forms could be supported concurrently for the following threshold:
  • Low frequency (pulling data once a day or week)
  • High volume (40,000 form submissions)
• [ ] Use findings to set the top tier for the token bucket and identify if this is enough to support 1% of forms clients

[srtalbot]

Research question: How many forms could we support at one time if they are conducting a bulk download of 40,000 form submissions per day. (with 2.01 actions per submission)?

Next steps: How might we adhere to this max capacity with our clients?

• Existing: Token bucket tiers
• Future: Setting expectations with clients on when to pull bulk downloads.