Restaurant Caching

Pyramid of caching
This platform is heavy featured with caching for the restaurants & venues. All caching is done in redis keeping the backend stateless. Let's start from level two caching which is distanceMatrix:

DistanceMatrix

DistanceMatrixService uses Spatial Caching for the api call with Google Distance Matrix API and this works when a different customer within 400m m radius makes a request. The caching is done from the userLocation key and I have rounded the lat and long to 3 decimal ex: 41.324,19.822". So similar users within 400m meters can get identical results. that we can map to our branchSummaryDto inside the mapper. If no error occurs, the method returns a RestaurantSummaryDTO. Time to live is 6 Days the reason is because the restaurants will not change the location very often (I implemented cache evict for this scenario) but what you might think is that user would never come online and Im saving stale data but this is not only for one user but multiple users in same area and also I calculated:
Tirana area = 41.3 km² Each cell = 0.4km × 0.4km = 0.16 km²

Unique zone keys = 41.3 / 0.16 = ~258 zones × 20 branches = 5,160 keys × 264B = ~1.36 MB
Completely negligible.
I only chach ids of the venues and not the actual information like name, address, phone number...etc. This is done to prevent execive calling to the google api matrix.

    // cache ~400m radius
    private String getUserLocationCacheKey(double lat, double lon) {
        int precision = 250;

        double latKey = Math.floor(lat * precision) / precision;
        double lonKey = Math.floor(lon * precision) / precision;

        return String.format("%.4f,%.4f", latKey, lonKey);
    }

DashboardRestaurants

This is the most aggresive caching and come before the distance matrix. This is cached for smallest time due to stale data in redis for long period of time and can fill up ram fast. So basically this cache the Page<RestaurantSummaryDTO> but with a wrapper serializablePage because without it it can't serialize. Example on Redis Insight.
cache key: dashboardRestaurants::delivery_time:0:41.353,19.749

PartValue	Meaning
cache name	dashboardRestaurants
sort	default
pageNumber	0
roundedLat	41.353
roundedLon	19.749
cache purpose	Rounded user location grid cell

Results

Before caching 43ms for the request
After the cache 16ms, approximate:260% faster (Not considering latency)