Friday, 9 December 2011

Determining rate limits in the AdTotum API

Writing the #AdTotum #API. Because the API is for internal use and use by academic modellers who work with us the Rate Limits will be issued on a case by case basis. Looking at the architecture though it looks as though internally caching the data makes a lot of sense (and in fact lots of the high demand data can be accesses straight out of RAM in the old time honored ways that I've been using over many, many applications. Typically in the AdTotum T745 model results are updated on a 60 second cycle which means that recommendations in the cache are effectively updated in real time. The cloud servers seem to be able to take 1M+ users on a single node (or micro-node) at about 5% load under NGINX and the API looks at though it will scale in a similar way.

No comments:

Post a Comment