To build great products you have to test, and we do all kinds of automated and not-so-automated (human) testing at Interactive Intelligence. The BAT (Big A** Test) is a weekly test we run to push scale improvements of our systems. These tests are not designed to verify accuracy, but to find breaking points by pushing systems beyond what normal customers/developers/test loads do.
Our testing continued last week with PureCloud. For those of you who don’t know, our PureCloud platform is a stateless and independently load-balanced set of microservices using Amazon Web Services. It is built for immediate and continuous delivery of new functionality and unsurpassed reliability, scalability and security.
Okay, back to the test. Here are some results for the telephony service:
- We placed about 4.8 million calls over four hours (333cps, calls per second) and approximately 2.55 million connected (177cps, 53 percent)
- During the peak three hours we placed about 4.1 million calls (380cps), with about 2.2 million connected (203cps, 53 percent)
- During the peak one hour we placed about 1.5 million calls (417cps), with about 761,000 connected (211cps, 51 percent)
- Net results: PureCloud can process 330 calls per second
This particular BAT test averaged 30-second long calls to maintain a high call rate with fewer Edge devices. PureCloud Edge is our hybrid appliance+cloud telephone service that also provides remote survivability across locations.
“We threw everything we had at this, and still everything just worked. It was impressive!”
– Adam Dyer, Cloud Apps Development, Interactive Intelligence
From Interactive Intelligence Software Architect Anthony Roach – “If we extrapolate this to 10 minute long calls, then this call rate would keep about 200,000 agents busy. If we assume two shifts of agents, then that would work out to be about 400,000!”
We’re not done yet – we’re continually improving performance and scaling in the ever-growing PureCloud platform.