Upload
apigee
View
622
Download
4
Embed Size (px)
DESCRIPTION
Check out these slides from I Love APIs 2014 to learn how architects and engineers are building products at Apigee, and how we incorporate open source technology for enterprise grade apps: Cassandra, Postgres, Netty, Jersey, jQuery, Bootstrap, Tomcat, Puppet, Hadoop, Couchbase and more.
Citation preview
We Built this City(Although not this one in particular…)Greg Brail, Gary Forrest, and Paul Kim
2
Overview
• A little bit about architecture• How we monitor at Apigee• Some real data
A Little Bit About Architecture
4
We try to stick to our principleswithin reason
5
Principle: Eventual Consistency
Consistency
Availability
Partition Tolerance
RDBM
SCassandra
Zookeeper
6
Eventual Consistency Leads to…
• Better availability– Support for multiple regions and data centers– No downtime due to hardware or network failures– No downtime to replace a component
• Better scalability– Add nodes without downtime– Scale horizontally
7
Single- versus Multi-Tenant
Tenant A
Tenant B
Tenant C
Tenant A
Tenant B
Tenant C
Single Tenant Multitenant
8
A Little of Each
Tenant A
Tenant B
Tenant C
Tenant A
Tenant B
Tenant C
Tenant D
Tenant E
Tenant D
Tenant E
Tenant A
Tenant B
Tenant C
Tenant D
Tenant E
Inbound Routing API Call Processing Data Storage
9
Everything Should Have an API
10
Everything Gets Deployed to the Cloud;Most Things Get Deployed On-Premises
How We Monitor at Apigee
12
Traditional Monitoring – information silos
AlertsCustom Checks
* Product names for illustrative purposes only. No endorsement by Apigee is intended or implied.
A different approach – standard alerts
13
standard alert
Custom Checks
Event Processor
* Product names for illustrative purposes only. No endorsement by Apigee is intended or implied.
A different approach – standard alerts
14
standard alert
Problem Details Host, service,
metricActions-to-take Graph Playbook Escalation contact
Custom Checks
Event Processor
* Product names for illustrative purposes only. No endorsement by Apigee is intended or implied.
15
Monitoring ArchitectureSystem Being
Monitored
Event Stream Processing
Server(Real Time Alerting)
Notification IM chat, ticketing system,
email groups, phone
Monitoring Agent
Visualization
Queryable Metrics (CloudWatch, API,
Logs)
MonitoringAgents
api-curl-agent
cloud-lb-metrics
16
Support Ticket History
17
P1 Incidents Caught by Monitoring
Some Real Data
19
Sports News API
08/29 08/30 08/31 09/01 09/02 09/03 09/04
When did the 2014 NFL season start?
20
Online Ticket Business
Traffic spikes before events
21
US Retailer
Labor Day Weekend
Daytime spikes and higher on weekends
22
Mobile Application Monitoring5-day traffic view 12-hour traffic view
Thank you