- We will group cities by proximity in regions.
- Depending on the geohash key length, we will need different number of geohash keys to cover the region. By default, we are using
5
with the length of 4.9km x 4.9km. - We expect ~100K active drivers in one region.
The amount of QPS and bandwidth we can expect are:
We expect ~100K location updates/writes per second Each location message consists of
- RideUuid 16 bytes
- PassengerUuid 16 bytes
- ClientUuid 16 bytes
- TripUuid 16 bytes
- Lat, Lon 16 bytes
- Timestamp 8 bytes
- Other metadata 32 bytes
So we get 100,000 * 120 = 120,00,000
or 12MB/s
(almost) bandwidth upload.
For peak our, we can expect to double it and assume 12 * 2 = 24MB/s
- Before starting a ride, we assume that the ride goes to the trip origin and starts the trip from there. If the ride is in a bit far away position, the travel point to trip origin is not recorded in the simulation. But endpoint to do so does exist to do so.