📖

README.md

@@ -22,7 +22,7 @@ click here for a complete schema. The data is generated by placing both drivers
 ## Matching
 The matching is heuristic and driver-centric:
 
-1. if a driver available / idle, scan for passenger nearby.
+1. If a driver available / idle, scan for passenger nearby.
 2. Dispatch to pick up the passenger.
 3. Switch status to 'On trip' to the destination. Along the way, unless cab is full, driver continuously scan another passenger nearby with common destinations, and re-route if necessary.
 4. Once arrived, drivers' status are set back to idle to pick up other passengers. Passengers removed after 2 hours later.
@@ -54,15 +54,17 @@ Data streams for passengers and drivers generated separately in python (Kafka pr
 Secor is used to saving all raw streams into Amazon S3 for later purposes (batch, re-play, forensics, or analytics).
 
 __Stream processing__
-ed in Spark Streaming with window 3 seconds, consuming data streams from both drivers and passengers. Every incoming message is subject to sanity check: e.g. driver's reported status is matched with the previous status, etc. to anticipate latency.
+
+Recieved JSON streams are processed and transformed in Spark Streaming with window 3 seconds, consuming data streams from both drivers and passengers. Every incoming message is subject to sanity check: e.g. driver's reported status is matched with the previous status, etc. to anticipate latency.
 
 __Sink__
 
 Elasticsearch is used as the buffer/transactional interface for the resulted messages. Elasticsearch is also called by Spark to enable assignments. Elasticsearch geo-location and boolean queries are leveraged. 
 
 __Output__
 
-The output is served as API by using Flask. Bootstrap2, jquery, and leaflet are used to prettify the output.
+The output is served as API by using Flask, output intended  mainly for dashboard and sent back to passanger and drivers.
+For Dashboard: Bootstrap2, jquery, and leaflet are used to prettify the output.
 
 # Infrastructure
 Hosted in Amazon S3 with 3 m3.large instances for Spark Processing and 4 m3. medium instances  other services (multitenant).