7. R
R
MP
MP
C*
C*
Region N
R
R
MP
MP
C*
C*
Region 1
Anatomy of an API call
7
Apigee
infrastructure
Incoming
app traffic
API proxies
and policies
Backend
targets
Valid requests
200
Invalid syntax
Wrong password
Expired token
Bad client requests
4xx
Proxy code
Policy violation
Callout errors
Failure
5xx
Memory
Bugs
Server down
Failure
5xx
Long latency
Timeouts
Server down
Failure
5xx
13. How do you measure latencies globally?
13
Edge of Network
Synthetic API calls
Washington DC - US
Oregon - US
Dublin - Ireland
Sydney - Australia
APIs
1.Measure response times
2.Validate response codes
3.Compare across regions
4.Alerts upon problems
API Health*
Quick Check-Up for your APIs
* Beta
23. Mitigate
• Rapid response to API traffic issues
– Administered by Ops teams without requiring proxy editing
– Remediation actions to temporarily restore API traffic
• Isolate offending parties
– App blocking – respond to rogue or compromised apps or keys
– Traffic rerouting – redirect inbound traffic to new location
• Graceful degradation
– Traffic throttling – reduce flow to backend systems
– Fail fast – reduce long time-outs
Traditional view of Availability based on uptime is flawed
System can be up but API calls could be really slow or failing
SLA = guarantee that API Success Rate > x%
API Success Rate = % API calls successfully handled
Manage risk and make commitments
API availability is critical to user experience
Unexpected API errors == potential lost revenues
Single metric measures quality of end-2-end API delivery network