Hotspotter is a microservice that receives keys that have been requested over time and stores counters using Redis.

Counters are stored in buckets for each hour, as a Redis sorted set (zset), and are aggregated upon request. Aggregation results are cached for a configurable time (default is 2 minutes) so that requests can be handled more efficiently. Given that this service is intended to be consumed from a load balancer, this cache should be tuned according to the expected responsiveness of the system to new hot keys.

See Documentation

wget http://download.redis.io/redis-stable.tar.gz
tar xvzf redis-stable.tar.gz
cd redis-stable
make

Just execute the "redis-server" script on the src:

redis-stable/src/redis-server

You can run the application on the default (dev) environment by executing:

sbt run

To specify another enviroment (such as prod), use the following command:

sbt '; set javaOptions += "-Denvironment=prod"; run'

The application will run on port 9290 by default.

You can post several key hits at a time:

POST /key-hits

with a body that follows the structure:

{
	"hits": [
		{ "key": "1234" },
		{ "key": "5678" }
	]
}

You call this service as "fire and forget". The keys will be processed asynchronously.

You can retrieve a list of the hotspots of the previous H hours (H is configurable):

GET /hotspots

The response will contain a list of the top N keys (N is configurable) sorted by request count descending:

{
	"keys": ["1234", "5678", "..."]
}

GET /hotspots/:key

Status: 204 No Content

Status: 404 Not Found

• Local cache of last hotspots to respond in case of a failure on Redis cluster
• Store a local oplog file and recover from it in case of a failure writing the keys to Redis.
• Performance/Stress tests using Gatling.
• Enable configuration of custom buckets instead of fixed hour buckets.
• Add a way to post summarized hit counts instead of having to repeat the keys.
• Figure out how to get rid of the annoying security pop-ups when running the tests with the embedded Redis.