I'm not so sure that low latency is the most important thing in each case. Personally, in our case, we wouldn't care - specifically we're only requesting (detailed) information per IP on a small set of specific events (i.e. login, signup) and additionally those calls to ipinfo/ipdetective would be async anyway. So perhaps there are two use cases here: 1. low latency, high volume). 2. Higher latency but lower volume or don't care about minimising latency to below 20ms.