Platform API update
by Udi h Bauman
1. IoT
1.1. millions of connections => billions of events
1.2. problem: finding the event that matters to the business
1.3. architecture
1.3.1. Jetty -> Kafka -> Storm
1.4. use cases
1.4.1. stream of events coming from vehicles, machines &c
1.4.2. processed & displayed in SFDC apps
1.4.3. expected scale is huge
2. Enterprise scale
2.1. API types
2.1.1. SOAP
2.1.1.1. same backend as REST
2.1.2. REST
2.1.3. bulk
2.1.3.1. 45M rec/hour
2.1.3.2. 5K batches limit
2.1.4. stream
2.1.4.1. feed of events - new results of a SOQL query
2.1.4.2. 1 sec latency
2.1.4.3. 1M events limit
2.1.4.3.1. for UE
2.2. API usage - getting near 2K BB records / month
2.3. Generic streaming (pilot)
2.3.1. enable developers to stream anything
2.4. Flexible limits
2.4.1. Monitor resource utilization, & relaxing limits if possible
2.4.2. pilot (request enablement)
2.5. API capacity allocation
2.5.1. allocate percentage of API quota to each connected app