Load Testing Notes
Nick M. starts out with a CLI tool - apache bench, but he recommends WRK
What url, how long, how many concurrently,
Slowly scale up the concurrency,
He also likes some cloud-based tools. Two that he likes: better free tier = loader.io
We may have access to paid BLITZ.IO
He suggests doing it directly at the endpoint since api.data.gov has rate limiting and throttling and even when you remove those, we still have others in place
About 50 concurrent from a single IP; 200/second
Does the Auto API allow caching? If so, api.data.gov could do.
Possibly see if logs.cloud.gov can allow
Failed requests = just plain bad
50 is pretty good - 100+ likely more than we’re likely to experience
500 ms under load is okay; without is not as good
1,2,3 second response time is pretty less ideal
Load Testing Notes Nick M. starts out with a CLI tool - apache bench, but he recommends WRK What url, how long, how many concurrently, Slowly scale up the concurrency, He also likes some cloud-based tools. Two that he likes: better free tier = loader.io We may have access to paid BLITZ.IO He suggests doing it directly at the endpoint since api.data.gov has rate limiting and throttling and even when you remove those, we still have others in place About 50 concurrent from a single IP; 200/second
Does the Auto API allow caching? If so, api.data.gov could do.
Possibly see if logs.cloud.gov can allow
ab -c 1 -n 10 "https://sec-autoapi.apps.cloud.gov/alepdc"
Failed requests = just plain bad 50 is pretty good - 100+ likely more than we’re likely to experience 500 ms under load is okay; without is not as good 1,2,3 second response time is pretty less ideal