What's the best way to conduct load or performance testing?
We've recently been asked, "We’re planning to scale up in transaction volume in the next few months, so we wanted to conduct some performance/load testing in order to validate our integration with Spreedly. Can we load test against the Spreedly Test Gateway? Are there any rate limits that we should observe? Is there anything else we should be thinking of?"
There are a couple things to keep in mind for load testing.
First of all, there is no such thing as a test environment at Spreedly; all environments are created equal and can accept both test and production transactions. When you run transactions against a test gateway, you’re talking to the same set of Spreedly servers you’re talking to when you use a production gateway.
Additionally, most of the time you spend waiting for the transaction to complete in a production context is the time that the gateway is processing the transaction (in other words, you’re not waiting on Spreedly). With the Spreedly Test Gateway, you’ll get nearly instantaneous responses, which are not representative of transacting against an actual gateway. With no “real-world” delay, the accuracy of the load test is greatly diminished, or potentially non-existent.
One suggested workaround from our Engineering Team that you might consider: for your load testing, if it might help to simply mock out Spreedly calls with a delay you’d expect from a given gateway (perhaps 2-5 seconds).
Finally, we understand that testing high loads before releasing a new feature or going into production is an important process for some organizations. However, it is crucial to remember that Spreedly reserves the right to throttle or ban testing that we deem to be excessive. We want to keep our core services up and running for everyone, and we prioritize production transactions over test ones. If you’re planning on performing high amounts of load testing through our test gateway, please contact Support to start the conversation.
Please sign in to leave a comment.
Comments
0 comments